While not exactly the same, I once got a call from a number I didn't recognize, and when I answered the phone it was a recording of my wife saying "Hello?". I no longer answer phone calls by saying "Hello", unless I know the caller.
I have a system that takes it one step further and both reduces the awkwardness and false-positive rate at the same time: I add the people that I know to the contacts on my phone. When a call comes in as a number instead of a name, I simply decline to pick it up. If it's not a spam call, they will either leave a voice message or send a text. If they do neither, then either it was a spam/scam call, or whatever they had to say probably wasn't that important in the first place. Win/win.
I've been doing this for a little over a decade and it hasn't let me down yet.
It's let me down a ton. Deliveries, contractors, maintenance people, doctor's offices with a last minute appointment available, and so forth. Fortunately never for a true emergency, but that's also something to keep in mind as well.
There are lots of things that people simply don't leave a voice mail or text because if they can't contact you immediately, there's no point. Or if the contractor can't get you on the phone, they'll just move onto the next home and skip work on yours that day or that whole week.
This is a specific example of what should be a much more general practice: having separate protocols for establishing an initial contact and establishing a communications session with an already existing contact. My email spam filter is based on this. It does a first-stage separation between email from people I've corresponded with in the past and everything else. That simple heuristic is enough to achieve >99% accuracy all by itself.
That's my approach as well, but I had the same number calling me for 3 weeks and I finally answered. It was my electric company, something had gone wrong with a payment.
They have my email address, they send me txts all the time, but apparently collections is still making phone calls. Had to be the dumbest thing I'd seen. Once I answered and found out the issue, I paid the bill properly, but I wonder how far it would have gone before they cut off my power, while they kept sending me emails and txts about things that have nothing to do with my bill.
I've had a disturbingly large number of repeat calls from people who absolutely refuse to leave a message. And it's always some recruiter who saw an opening on indeed or somewhere and thinks the resume I updated 5 years ago is a good match.
The problem is that if I'm getting repeated calls from an unrecognized number, I'm assuming my wife, my kids, or my parents are in an ambulance, so I have to drop everything and answer.
As a rule of thumb, if I get a one-off call that doesn't leave a message, I'll search my email inbox for that number, as they've probably contacted me separately. However, one time, I got called 5 times in 90 minutes, with the only message being 23 seconds of silence, and an email I hadn't even read yet (searching the number brought up the email). I sent an angry email that amounted to "you have told me how you AND YOUR CLIENTS treat prospective employees' time. I will never apply to any job you suggest, even independently of you. Stop calling"
Many of us are in situations where we get calls from various people we haven't had contact before (nurse at the child's school, parent's doctor, there's a lot of them) that should be answered immediately; waiting until later to listen to the message could have significant impacts. Some of the calls (injured child) could require immediate contact and, if not answered, could result in other issues.
This 100%. iPhones have a feature to do this automatically. It doesn’t even ring, and goes straight to voicemail if they’re not in your contacts. It’s so freeing!
One major flaw in this, at least for me: Dr's offices. They love to dial from a gazillion random numbers, and for privacy reasons they often leave no message or a very vague and concerning "Call us when you get this" sort of thing.
I have a different system. I pick up the phone, listen to them for a bit, tell them "please wait while I get my credit card number", and then I just walk away with the connection still open.
It's great when it works, but when my mom was in the hospital and they needed to reach me, I got burned by this big time and don't do it anymore. It's too easy to miss a call that could literally be life and death (my mom is better now).
I do exactly this but take it even one step further. My actual (primary) phone number is only ever given out to humans. I have a second Google Voice phone number that I give out to machines (e.g. online shopping that "requires" a phone number that will eventually be leaked).
I have a child, he has a phone but his battery might go empty, or the phone is lost or broken, he has my number written down and I instruct him to call me from a colleague or a stranger. Maybe my case is special since my son has some health issues so I really want to know immediately if something happened.
This kind of problem needs to be solved at the root cause, say if the phone companies could be made to pay a bit when you get spammed and forced to recover their costs from the spammers the issue would be solved, now if they profit the issue will get larger and alrger.
This method unfortunately falls apart if you get a phone call from a hospital. They'll leave you a voice message, but when you call the same number back you'll get the front desk instead of the doctor who left you the message. They'll patch you through to the ward your Dad's in, but they won't be able to give out any information over the phone, so you'll need to wait for the doctor to call you back. They're out doing their rounds at the moment, but they'll get back to you as soon as they can.
Yup, same. I'll make an exception if I'm expecting an important call but aren't sure of what number it's going to come from. This is rare enough that it doesn't bother me much. And now that some calls are SHAKEN/STIR-verified, with a caller ID, I can often have good confidence before I pick up that it's actually the call I'm waiting for.
I do this too, but I also remember that I'm doing this from a situation of privilege, where I mostly don't have to wait for calls that could be life changing (ex: old-school HR calling back for a new job).
On Pixel phones (or was it Google Fi? can't remember), this is automatic. If it's not someone in my contact list already, known spam gets auto blocked and everyone else gets redirects to the voice assistant that takes a message and transcribes it. Cuts down on spam like 99% for me.
I had an iPhone for a few months and the spam was so bad, even with the third party spam blockers. I switched back to Android shortly after.
I do the same thing usually. If I do pick up an unknown number because I am expecting something, I usually press speaker and mute and just wait. If it's a person, I'll get an awkward Hello? And if it's an auto dialer usually I get nothing or the waterdrop beep and drop either way.
I do a thing where I answer and just dont say anything (ensuring my enviornment is silent) for like 20+ seconds.... they hang up and I block number. (The bot thinks its a dead num and I dont get calls again.
If your car gets stolen, and the police find it, they will call you from a phone number that's not in your contacts. If you don't pick up, you won't realize that your stolen car has been recovered a couple miles from your house, and if you show up there in 30 minutes you can drive it back home, but if you don't, the police will send it to a towing yard, which will require you to go through 24 hours of paperwork with the police to obtain a release and then pay the towing yard $1,000+ to tow and store your car.
If you live in an area of low crime, though, maybe it'll be fine not to answer phone calls from numbers that aren't in your phone.
Precisely, I give zero information. If I do pick up once in a blue moon, I pause for 3-5 seconds to give a chance for the human to start (if it isn't a bot).
I have a Pixel phone and a Google bot can answer the phone for me. It transcribes the conversations on my phone in real-time, and I can push a few buttons to tell to bot what to say--things like "tell me more", or "please tell me why you're calling".
If the entity calling gives an explanation I care about, then I can press a button and the bot says "thanks, connecting you now" and then I can say "hello" with my own voice and have a normal conversation. I think most people think it's just a fancy answering machine, they don't realize I'm controlling it.
Voice calls are on the decline anyway, but I think it's becoming possible to have a very sophisticated AI secretary answer calls for you, even beyond what I've explained Google is doing. Imagine being able to give your LLM phone secretary a prompt and it would answer calls for you. You could tell it something like "the snowblower I listed in the classifieds is already sold" and maybe it could automatically resolve some calls or text messages for you.
The phone system has gotten so bad these days that a lot of the time the pausing for 3-5 seconds isn’t voluntary - it just doesn’t connect the call properly. The most basic hundred year old regular phone call is too much to handle for modern systems I suppose
same, but now a lot of callers whom i would like to speak with -- e.g. my insurance company -- just hang up before greeting me (because they think my phone's broken?). but then if i screen everyone via voicemail instead, a different (but overlapping) portion of callers refuse to leave messages. it's like everyone's given up on using the POTS outside of their immediate social circle, and the few people/businesses who still do are either malicious, or are just going through the motions.
thanks spammers. and thanks FCC for sitting idly over the decades and letting the spammers ruin it. weird time to finally put your foot down, but sure, okay.
I've been getting these calls where nobody says anything for like 3 minutes then someone says Hello. My paranoid mind thinks they are trying to record my voice to use AI to impersonate me.
> My paranoid mind thinks they are trying to record my voice to use AI to impersonate me.
You're not paranoid, banks, the Minnesota Attorney General and the FCC have been warning about scammers recording even as simple as a "yes" to use in their scams [1][2][3], although actual evidence has been scarce to say the least [4].
My thought has been that they're listening for background sounds to try to beef up the advertising profile they have on me. Maybe there is some super sketchy ad-tech company putting beacons that emit a QR-like UUID audio signature in the frequencies near the top and bottom of the range that gets transmitted by cell phones, and ringing you up from a robo-dialer and listening for the beacons tells them where you are.
FWIW, I get these, too. All unknown numbers go straight to voicemail, which auto-transcribes, so I just see "Hello... hello..." in the transcription and hit delete. No idea what it's about.
I got a call sort of like that, it was bizarre. A person claiming to be a Comcast rep called, introduced themselves, asked if I was me, and then immediately hung up as soon as I made a noise.
It is possible they just hung up because I was already a little skeptical and feeling cagey, so didn’t give an enthusiastic “yeah that’s me.”
Anyway, I’ve never been called for something that benefits me. So, hopefully every company that depends on cold-calling will go out of business soon as everyone younger than, like, halfway through gen X doesn’t pick up their phone anymore.
Sorry for the breach of phone etiquette but I am on the same page here - the caller needs to speak first so I can tell whether they're a real person or not. If it's an automated system I'm happy to remain silent in the hope that they don't realize my phone number isn't another automated system.
I guess you'll end up confusing a lot of people since it's exactly backwards from the normal handshake.
Although you're not alone, most of the time when I call customer support and it's an overseas call center, I have to say Hello 2-3 times before the person on the other end acknowledges my existence. I guess they don't realize that I can hear all of their background noise before they talk.
Maybe if I just placed a delivery order I will answer for an unknown local number. Beyond that, leave a message at the beep and maybe I'll check it in a few days.
When you’re dealing with contractors and whatever for house stuff, yeah you kinda need to answer the phone for long stretches of time. Same if you have kids (I don’t), you need to be receptive. Yes yes I am incredibly aware that people can leave voicemails and send text messages, but many out there won’t do it, from real experience, especially those outside of the tech bubble.
I have a friend who would always answer the phone with a robotic monotone "READY" like a C64 BASIC prompt. It made people think he was a robot, and confused the real robots.
I have gotten into the habit of answering the phone in the Graham-Bell/Mr. Burns way by answering "Ahoy Hoy" whenever I get a number that I don't recognize. I figure that that's not going to be as useful for any training purposes, and is also pretty inoffensive, so even if I don't get a robot then it won't offend anyone.
> I figure that that's not going to be as useful for any training purposes
Um what? Why? It's just as much a sample of your voice, and if it's what you usually say on the phone then a recording of it will... sound like it's you on the phone.
I just don't care. It's not like they can train a bot to convincingly speak like me from just one word. And if they can, the game is already over and we've all lost.
That said, I don't answer suspicious numbers and I won't move past "hello" until the caller identifies themselves.
That's going to be a major and widespread issue very soon.
Unfortunately, rulings such as this FCC's are ineffective to prevent it. If someone is already committing fraud, they obviously won't care if it's illegal to use an AI-generated voice.
If I immediately hear sound from the caller it's usually a valid call. If I wait several seconds and it's just quiet, it's an automatic dialer waiting for a voice response. I found it highly effective at weeding out spam calls.
Wonder how many secs of voice you need to replicate one. You can call a number programmatically, ask something silly. record the response and then recreate the voice. I can imagine one can do much harm. Like calling the voice's boss and tell him you fell in love with his wife and now resign.
This is why I love Google's new AI phone call screening feature. Some people get spooked by it and hang up, and sometimes spam calls get through via exploits like calling twice within a short time or somehow bypassing with a weird spoofed number (only happened 1-2 times so far)
I only answer the phone with "Who's calling?". If I don't want to talk to them, they get "this is his assistant, he's not available". If it sounds even slightly like a canned voice it gets hung up on.
I was once told that some automated dialing systems will listen for, and hang up/flag the number as another automated system, if you wait four seconds, say hello very clearly, and then say nothing else.
There are a series of gates. At the end is the scam. Each gate is designed to filter out those who will reach the end and not fall for the scam. Or in other words, by the time you are making the scam pitch, the scam is already done, because you know by then it will work.
The calls are just one of the early gates, as someone screening your call is likely not to fall for the eventual scam.
The gates don’t have to be clever for this to work. There merely has to be enough people that you are going to find that 0.1% who will fall for it.
I think it's about proof that the number puts them in touch with a real person. I suspect if the robocall gets enough engagement they'll even put an actual scammer on their end.
Yep, I don’t say hello anymore either, if I don’t recognize the number. Makes things awkward sometimes, but this is the dogshit awful world we live in.
I think it is important to note that the legal principle that allows the FCC to make rulings like this is called Chevron Deference, and many consider it to be under attack.
This thread wildly misunderstands "chevron deference". "Ending chevron deference" does not somehow throw us into a Mad Max anarchic hellscape where agencies cannot actually do anything, because there is always some standard for what administrative rulemaking is permissible. There is a broader question of how much leeway they have, but clarifying that AI generated voices count as "artificial" under the statute barely requires a regulation, any more than they need one to say "hit in the head with a computer" constitutes an "assault".
The problem with your argument is that, for decades, congress has been passing and failing to update laws under the understanding that the courts would apply Chevron deference.
If the courts decide to get rid of that, they're intentionally misinterpreting the laws that congress has passed over that time. They're also effectively rewriting a large fraction of US law, despite the fact that the constitution is carefully designed to prevent such a small group of (unelected or elected) people from modifying US law that quickly, and without safe guards.
The current Supreme Court has repeatedly undermined separation of powers, and they're explicitly doing so against the wishes of the electorate. Their behavior is fundamentally undemocratic.
Imagine the following: The FCC fines a company for using AI-generated voices in robocalls. That company appeals the fine. With Chevron intact, the court would need to defer to the FCC's interpretation of the TCPA and dismiss the appeal. With Chevron overturned, the court would be able to advocate for their own interpretation of the TCPA. A favorable judge could just claim textualism, and insist that the TCPA does not apply because it does not explicitly use the word AI. Then it is a slippery slope of forum shopping and companies moving their operations to districts with sympathetic judges.
Even if it was unclear, ending Chevron deference wouldn’t say “the agency can no longer make these policy interpretations.” It just means that a court ought to test whether that interpretation is in compliance with the law, when that comes up in a dispute (which is something that courts are in the business of in many other areas) more so than simply deferring to the agency’s expertise on the law.
(If you look at the original Chevron decision, they were much more interested in trying to get out of the “understand and make determinations about complex environmental issues” business anyway, more so than the “understand the law” business.)
Postscript: For your next unfairly downvoted reply I recommend that you explain to someone Citizens United was actually a nonprofit trying to air a movie on cable television and was fighting the FEC over it. (Total hackjob of an organization, mind you. But core political speech.) Some facts are unpopular.
Chevron deference would come into play if the FCC tried to say that a test-tube baby was an artificial agent. I support ending the doctrine, because the shadow laws are strong and bad.
This is great that this is line of comments are under an article about banning something most people here would like to see banned. That is in fact doing something good, unless I guess you're on the side of robocalls. Perhaps choose to make this argument in another thread, it'd be far more convincing.
I've only read your link there, but aren't you mischaracterizing it? The Chevron doctrine isn't what allows agencies like the FCC to make rulings like this, it's what protects their decisions from being overruled by the courts. That is, even if all the justices privately agreed the agency's interpretation had issues, they'd still defer to it. But without Chevron, in that case, they could overrule it.
In this case, considering AI-generated voices "artificial" for the purposes of applying a law seems obvious enough to me that I don't think the Chevron doctrine would apply, personally.
> The Chevron doctrine isn't what allows agencies like the FCC to make rulings like this, it's what protects their decisions from being overruled by the courts.
Yes and it's in cases where a law gives authority and expectations to an agency. In the past, it was left up to experienced and qualified agency specialists to work out how best to implement it because 1) it's their job and 2) because Congress knows it can't write every possible contingency into a law.
Chevron supports this. The SCotUS case is brought by folks who want to shift that determination from agency specialists to judges who don't have the related experience or qualifications. It would effectively allow endless monkey wrenches to be thrown into the oversight process by corporations who aren't keen on oversight.
I agree with you on the importance of Chevron deference, but I can't see any court getting to the second step of Chevron with this particular ruling, so no deference would be required. The legislation bans "artificial or prerecorded voices"; AI agents are by definition artificial.
> "Congress should have used more precise language rather than deferring to the supposed "expertise" of members of the administration in order to establish the artificiality of AI"
- SCOTUS, in a judgement not yet issued or rendered (and thus currently wholly imagined by me).
Congress says: "Hey agency! You're the experts. Figure out and enforce the policy details."
The agency: we have determined that this action by company X is against our policies.
The courts: Congress said that the agency decides the policy. Even if we think an action is inside policy, the agency has Congressional authority to change the policy to put the contested action firmly outside policy.
The company should therefore lobby Congress to regulate the agency. Maybe you could make a case about retroactive or post-facto laws, but I suspect the company is not usually claiming that they abide by the letter of the policy, but that the policy is outside the agency's powers.
It will cause chaos and disaster if congress has to make regulations for every little thing. Congress is so divided the result of Chevron reversal is that huge numbers of usefully regulated utilities, companies, etc will be unregulated. It also doesn't make sense for congress to spend all their time writing regulations, they'd get even less done. Congress can barely pass a budget shortly before the previous budget year ends.
Ending the ability of federal agencies to write useful regulations means unregulated spam robocalls! It's the dream of Elon Musk and Peter Thiel. Rich people are unbounded. They would say we don't need regulations about food safety written by those ninnys in the federal government.
Yes. Heavens forfend if they had to do the job they asked to be elected to do.
In short, good. How many here can even map the entire list of all the agencies and corresponding rules, recommendations, and guidance that has the weight of law.
<< It will cause chaos and disaster if congress has to make regulations for every little thing.
Free people pull in all sorts of directions. Its going to be ok.
Congress should have gotten off their hands and written something by now, same with Crypto legislation. “Chevron Deference” breeds tyranny through legislative apathy
No, Chevron deference breeds sanity. it would be insane to think that every little detail of complex regulatory structure must be outlined specifically in legislation in order for it to be valid. For example, legislation gives the EPA the power to regulate waterways. Chevron deference allows the EPA to use its expertise to write rules that say you can't dump benzine into the river. Without Chevron deference, someone who wants to dump benzine in the river could challenge the EPA saying that the law doesn't specifically say you can't dump benzine in the river. Imagine relying on our elected officials to come up with a list of what is and isn't considered toxic.
>Congress should have gotten off their hands and written something by now, same with Crypto legislation. “Chevron Deference” breeds tyranny through legislative apathy
It would be literally impossible for congress to rule on every nuanced thing that Chevron allows agencies to do. Saying "congress should take care of it" shows either an intentional disregard for the roles agencies and their experts play, or a complete misunderstanding of the power it grants to federal agencies.
"It breeds tyranny" is absolutely ridiculous. When agencies rule in a manner people find unjust, they sue and win or lose in a court of law based on the content of the policy. It also gives congress a chance to rule on "big ticket" things that do need addressing without causing an absolute standstill having to rule on something as mundane as what the legal weight and length limit should be each season for catching a salmon from federal land in Montana.
From a practical point of view, it's hard to say whether Congress would make better or worse decisions, and it's probably good that the government can make decisions about new technologies while Congress is mostly dysfunctional.
Maybe the thing that guards against tyranny is that Congress can override them (by passing a law) if regulators screw up badly enough?
At least, in theory.
Just like, in theory, the people could elect a better Congress.
Congress did act. They passed the TCPA in 1991 knowing full well that Chevron deference would allow the FCC to tweak their interpretation of the law as facts change. Congress doesn't want to have to micromanage things like this. If they did they would write the laws in a way that prevents situations where Chevron comes into play. And anyway, getting rid of Chevron would transfer the agencies powers to the courts, not congress.
Have you listened to our congresspeople? Nothing they do or say suggests to me that they have the capabilities to legislate effectively on technical matters, be they AI, Pollution or Food Safety.
We have departments that have traditionally been staffed with SMEs to make these rulings and decisions on behalf of congress, who legislates their existence and budget.
Our country is falling apart because of the current level of congressional ineptitude. One party refuses to support important legislation they specifically asked for because it may give the opposition party a positive news article.
Wishing the Congress had to study and pass legislation for all enforcement and regulation of society is tantamount to accelerationism.
there is simply no way for congress to enact every regulation. This is all a power grab for corporations bankrolling republican judges and congress critters to be able to ignore any regulations they want in order to make a few more bucks.
I think the current SCOTUS thrives on chaos (6 of 9 members anyway) and Chevron will go down in flames just like Roe. This is the modern "conservative" party.
Glad to see Chevron Deference at the top here. Basically, the FCC can’t “rule” they can “dictate” and this isn’t a power explicitly granted by congress. It’s some made up judicial rules that say these federal agencies can do it
The controlling legislation here, the Telephone Consumer Protection Act of 1991, prohibits initiating "any telephone call to any residential telephone line using an artificial or prerecorded voice to deliver a message without the prior express consent of the called party" (I got this quote directly from the FCC ruling). The legislation does not define "artificial or prerecorded voice". The FCC here is stating that they interpret "artificial voice" as including interactive AI voice agents, which did not exist in 1991. Do you think this is an unreasonable interpretation? Or should Congress be required to list exactly what technologies are prohibited in this context and update that list every time something new comes around?
Chevron Reference is the idea that when a statute is ambiguous the agencies can interpret it according to their expert opinion.
The alternative is requiring Congress to write every single rule explicitly and pass a law adapting to any change in circumstance or technology. In practice this means "no regulation" because Congress is pretty slow and adding more detail would only make them slower.
Executive agencies are granted authority by the legislature. The legislature can at any time make additional legislation overriding or limiting specific actions taken by executive agencies. It isn't made up.
Nonsense. The law in question explicitly grants the FCC the right to make this determination via regulation.
> The Commission shall prescribe regulations to implement the requirements of this subsection. In implementing the requirements of this subsection, the Commission — (A) shall consider prescribing regulations to allow businesses to avoid receiving calls made using an artificial or prerecorded voice to which they have not given their prior express consent; […]
Chevron deference is about whose interpretation governs when a law is ambiguous; that’s not even close to being the case here.
Who will think of the poor corporations and their armies of on-retainer lawyers?
Of course government is incompetent and can't be reasonable in regulation? Is that the idea? How dare these corporations not be given minutely detailed regulations that they can easily tear apart to pollute to their convenience? You mean you want REASON in government and regulation?
It would be more correct to say that they have officially interpreted a current law (the Telephone Consumer Protection Act) to clarify that AI-Generated voices in robocalls violate that law, which seems reasonable.
Activist? You clearly don't have the same definition of activist that I do. Half the problem with these sorts of conversations is there is no agreement on definitions.
Please don't interpret my comment to mean that Supreme Court decisions can't be criticized, I just don't find the "activist" accusation to be particularly insightful.
I think it's pretty unlikely that the Chevron doctrine would be overturned completely. The specifics of the case before the Court involve a case where the NMFS has interpreted a fisheries act to require fishers to pay the salaries of government monitors, simply because the act does not specify who should pay the salaries. The more reasonable objection is whether "reasonable interpretation" under Chevron should be limited to prevent the creation of affirmative powers out of thin air. As Wikipedia puts it:
> Whether the Court should overrule Chevron or at least clarify that statutory silence concerning controversial powers expressly but narrowly granted elsewhere in the statute does not constitute an ambiguity requiring deference to the agency.
The initial "overrule Chevron" seems like a DITF [1] and the latter is probably what the plaintiffs are hoping to achieve. Granted, I find it hard to trust a Court that overruled PP v. Casey, but most of this Court's other rulings, at least, have not been as extreme.
Usually I challenge people when they call a court stacked or activist, because it’s just so rarely true: this is as close as you’ll (hopefully) see to a 1-bit high court.
It’s the masterpiece, the magnum opus of the Magnus of parliamentary politics. Nicollo Machiavelli doesn’t have shit on Mitch McConnell at that chess game.
I’m pretty indifferent to which color bumper-sticker late capitalism is sporting when it pushes the newest round of formerly “looking forward to better” working people below the waterline, it’s not a partisan thing.
The other team have plenty who match Mitch on evilness, but zero on skill.
The old Bell companies are largely already in compliance with SHAKEN/STIR. It is mostly smaller shady companies that are not, because they know their customers don't want them to comply.
Better to abandon that technology all together (for normal phone calls). It should be used exclusively for emergency calls and similarly vital functions. Let everything else operate over cell networks and require explicit opt-ins before party A can call party B.
What should be done is something else entirely. Apple and Google should offer, as part of their standard software, a personal "phone robot". When you get a new phone, you spend 15 minutes recording various phrases, and from that point on you just have the robot answer for you.
When the robot talks to these spammers and telemarketers, it will try to keep them on the phone as long as possible. A minute would be good, 10 minutes would be better. As the spammers tried to avoid this, Apple and Google could improve the robots to counter.
And, within a few months of this, at most, that industry would just be dead. It can't afford to spend a half hour on each call trying to determine if they've got a real live knucklehead who will start sending cash to Nigerian princes, or just bad software tricking operators who don't speak English as a first language. Their margins would drop, their need for more sophisticated AI to try to determine if they were talking to a real person or not would skyrocket, etc. It just wouldn't be economically viable to continue.
Computer time isn't that expensive; I'm relatively certain that the calls I get are either fully driven by voice recognition, or by someone in the third world or in prison, pressing buttons that activate pre-recorded statements by a script.
The former is cheap enough that yes, they would engage for 15 minutes. The latter are smart enough to understand what's going on so that they'd hang up.
cryptographic signatures are going to have to start becoming necessary for all kinds of things, like even your average JPG image, otherwise nobody can tell what is "fake" or not, court evidence will start to become useless.
> The FCC announced the unanimous adoption of a Declaratory Ruling that recognizes calls made with AI-generated voices are "artificial" under the Telephone Consumer Protection Act (TCPA).
So illegal in the sense that artificial robocalls are already illegal, then.
You can't possibly expect Congress to give executive branch agencies enough money to do a bare-minimum job of enforcing the laws Congress passes. Especially when there are political donors making sure that we deregulate things that society wants controlled so that they can rent-collect.
It seems like they're targeting the symptom instead of the problem.
One of the biggest problems with robocalls is that it's really impossible to know who's calling you, and that SPAM reporting tools don't have much teeth.
IE, when I have an incoming call, I should be able to see who's liable for the call. IE, "[phone number] is registered to [Person or corporation]", and that reports of spam should impede that party's ability to use the phone network.
As long as 1% or more of voters in Pennsylvania keep voting based on whomever talked to them last; and as long as Super PACs can continue to receive unlimited anonymous money; no media channel will be legally restricted from spamming people. Phone spam is too effective politically.
I don't think there's much evidence to suggest that robocalls produce material swings in elections at all, let alone 1%, a number commonly attributed to all campaign GOTV efforts put together.
This was my thought too. While I do think going after this kind of scam is a good first step, I don't see overseas operators not using this any less. Most spam calls I get don't follow the do not call list, why would they follow this either?
I think the FCC needs to step up and have a hard deadline for STIR/SHAKEN with fines for operators who don't comply. That is the only way, IMHO, that the VOIP operators will take it seriously.
I think "SHAKEN/STIR" is supposed to fix this long term. I'm not sure why it's taking so long, but I believe phones will already indicate if the phone call has a verified caller id. Probably next step is to just block any non-verified caller. I'm assuming there's just a lot of migration work to happen.
I would say that money is the root of the problem. I think that most VOIP providers don't want to loose out on unencrypted traffic (both legitimate and spam).
Also, why do I seem to always get spam from a few providers? And why aren't we holding them accountable?
> Signed traffic between Tier-1 carriers increased to 85% in 2023
We're getting there, just not soon enough. The whole world will have transitioned to never answering their phone before this actually is fully enforced.
It's already possible to lookup the carrier of a number, and I'd love for the ability to be listed under their location on the incoming call screen. Makes a big difference if the call is coming from T-Mobile or some company you've never heard of.
Make it an option. I should be able to block my number from the receiver of the call if I choose. The receiver should be notified the number is blocked and can choose accordingly. The fact that numbers can be spoofed is what should be illegal. Any company making calls should have to identify themselves to the person receiving the call.
Do most people actually care about being able to place phone calls and be anonymous in 2024? If I call someone it's either someone who has my number already or someone who is going to ask who it is (like a business) and I'm going to tell them who I am.
When I visited the FCC many years ago, one fo the reasons they give for allowing anonymous calls is the the protection of domestic violence victims. Eg they may need to call their abuser to talk about child support payments. They shouldn't need to reveal too much information away, particularly if it could be used to find their address (eg a phone number)
I think what I would is a level playing field. If I get a call like that I cannot trace, I would expect that I should be able to do the same. If I am held to a standard that is not conducive to privacy, so should the person on the other side of that call.
But.. there is money on the line. Clearly, money from telemarketers/scammers/whoever is using this tech is enough to make telecoms hesitate from actually doing something about it.
Twilio had some strict policies introduced that I think were industry wise for USA. Basically all voip numbers had to go through thorough checks, which even our legitimate company failed (go figure). So as long as all companies like Twilio introduce those checks then spam calls should dramatically decrease. I thought it was already the case for USA?
It is maddening that the companies that provide the service appear to have thrown up their hands & pretend that they have no idea how they could possibly prevent spoofed numbers. Imagine if this was this easy to spoof IP #s. Perhaps it is.
I thought people's behavior these days was to ignore calls from numbers they don't know and let the phone screen it. I don't ever have problems with unknown numbers or SPAM calls on my Pixel
That's also the issue with swatting and fake calls to 911. When investigators trace it they'll hit a VOIP provider and it becomes near impossible to take it any further.
Be careful what you wish for. No reason why governments might decide they want the same thing for the Internet and domain names. Requiring a license to own domains… who are we kidding, they’d do it for the tax revenue.
> It seems like they're targeting the symptom instead of the problem.
I believe this is a quickly adopted band-aid in response to the recent political scam calls that pretended to be President Biden telling voters to skip voting in the primary.
Does this ruling make IVR systems illegal, too? I applaud the effort because this really could curb a lot of spam, but I am curious because AI generated voices in phone calls are already ubiquitous and have been for decades. Do they have a specific line they're drawing on quality of the voice?
The ruling specifically only applies to the initiator of the call. IVR is not covered. Automated calls are also permitted with consent (for instance, if you sign up for notifications for filled prescriptions or backordered library books). It has nothing to do with the quality of the voice -- prerecorded voices are banned too.
So is this going to be another clusterfuck like 10DLC? I am glad our company stuck with our guts and intentionally decided not to go outbound, but I almost feel bad for the startups that were banking on full outbound.
"Artificial" voices in telephone calls have existed since 1971. That is when the Votrax speech synthesis device was first developed by a company known at the time as the Federal Screw Works. The engineering was done by Richard Gagnon.
Enforcement is difficult, but tracking complaints back to the source telecom / source customer and taking them to court, generally.
Automated voice messages were already restricted, this ruling just affirms that AI generated voices fit the categorization of automated voice messages.
Here's some relevant text from the ruling:
> II. BACKGROUND
> 3. The TCPA protects consumers from unwanted calls made using an artificial or prerecorded voice. See 47 U.S.C. § 227(b)(1).
> In relevant part, the TCPA prohibits initiating “any telephone call to any residential telephone line using an artificial or prerecorded voice to deliver a message without the prior express consent of the called party” unless a statutory exception applies or the call is “exempted by rule or order by the Commission under [section 227(b)(2)(B)].” 47 U.S.C. § 227(b)(1)(B). The TCPA does not define the terms “artificial” or “prerecorded voice.”
and later
> III. DISCUSSION
> 5. Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice.
> tracking complaints...and taking them to court, generally
Incredibly prejudiced judicial procedure, given the power, size, globalization, and ease of automated calling systems vs the normal people they most affect. Multiplied by an already burdened court system.
> Automated voice messages were already restricted, this ruling just affirms that AI generated voices fit the categorization of automated voice messages.
This is helpful. This isn't a tip-of-the-spear ruling, then, just something that affirms another ruling. But regardless, it sounds easy but in fact necessitates a huge amount of burden.
Some people record their calls. Businesses often have to per compliance in most direct to consumer sales situations. From the recording, if not algorithmically, a court of law could easily determine an AI voice case by case.
A real call center would have a record of which employee made which calls when. The court subpoenas those records and the phone company's records. If they don't match, there are problems. Unless the company wants to commit perjury by inventing fake employees and call records.
For now this could be seen as an incentive for TTS solution providers - build a product that is hard to distinguish from an actual human calling. In many cases the results are already convincing.
And what about the future. Please scan your retina to initialize the phone call? Please solve a captcha to start a phone call? Your workplace registered 12948230 calls in the last 24 hours, but employs only 3 workers registered as humans, pay fine now? Interesting times.
They describe this as giving "State Attorneys General across the country new tools to go after bad actors behind these nefarious robocalls." The way that I read that is that there are these scams out there that states are already trying to bring lawsuits against, and this simply makes their job a bit easier in some of the cases they're ALREADY bringing.
An antispam idea in bitcoin circles is to require payment to open an email from an unknown source. So if I want to send you an advertisement, it will only reach you if I add a payment invoice that meets your threshold. It makes spam costly and forces advertisers to focus on a narrower range of ads to people who more likely want the product.
Being forced to interact with city government, state benefits, hospital systems, courts, police, and especially probation officers, all of whom are known to block or obfuscate their number even though missing their call could cause you no end of trouble, would help disabuse you of your smug solipsism.
And most services including utilities if you want to reach a human. It fucking sucks because English is my second language, I can read and type fine but if I have to talk on the phone I’m screwed, maybe both of us could understand 70% what the other party says
I’ve submitted several ideas to Apple over the years. One of them actually made its way into iOS, which is the silence unknown callers option. I’m very happy about that. Before they added it, I tried to implement it with the existing feature set by setting my default ring to silent, then adding a custom ring to all my contacts. It was a pain, but it technically worked.
One thing I really noticed was the dramatic drop in call volume once I stopped answering calls. Once I stopped answering, they stopped trying to call. People are basically being trained not to answer the phone.
That's one of the reasons I've not been huge on the recent (within the last several years) increase in "scam baiters" and stuff. As much as it does waste the time of the scammer and as helpful as some of the big ones are, normal people who do it are having little effect and ultimately just putting themselves on more and more lists.
> Callers who use AI technology must get prior consent from the people they are calling, the FCC said.
The text of the ruling says "prior express consent" instead of unsolicited. That seems clear, but I wonder whether it is in practice. Is the one of those things where, by signing up for website A and agreeing to their terms by clicking a checkbox, I am agreeing to allow my phone number to be called by robits from companies B-Z, because of some line buried in the middle of the legal text I didn't read? I.e. "The User consents to contact for any purpose by Website A and our partners", and a partner is defined as anybody who buys their contact list from them?
That is a case where the nature of T&Cs and end-user agreements makes the words "express" and "consent" more abiguous than they ought to be, since they rarely match anyone's definitions except the law's.
This came up on a thread the other day, and I think a good counterpoint is emergency evacuation orders for the elderly. My mom doesn’t use a computer, cell phone, tablet, etc, and a robocall to her land line would be the only way to notify her.
Couldn’t agree more. For most people robocalls are an annoyance, but for millions of aging seniors they are a direct form of elder abuse. The amount of confusion, fear, and actual financial ruin I’ve had to deal with with family members makes me wonder how it’s had been legal for so long
BMO uses virtual agents that impersonate humans in Canada to follow up on credit card promotions. I’ll give them credit for being realistic, I’d have fallen for it, except what I do for a living. I test them by saying:
“The maze isn’t meant for you.” (Westworld)
Cover is blown immediately. It awkwardly says it didn’t get that, then it agrees with me and tries to move on with prodding me to accept the credit card blathering on about the benefits.
Banks doing this is an exceptionally bad idea. It’s one thing to Robo call and be clear your virtual agent (though bad - and I like the idea of it being illegal), it is extraordinarily creepy and offensive to impersonate a human.
Does this also ban generated voices when they self identify as such? IMHO, if someone is not trying to deceive, it should be allowed. E.g. if the call starts out as "this is ai generated voice from xyz, ____". There are likely useful use cases for that.
> Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice. Therefore, callers must obtain prior express consent from the called party before making a call that utilizes artificial or prerecorded voice simulated or generated through AI technology.
So that disclosure won't work, unless (IANAL) you have a checkbox in your signup flow that says "Yes, I consent to allowing voices generated by AI call me."
A useful use case for the unsolicited caller. I don't believe there is a single useful use case for an unsolicited robocaller for the receiver to begin with, regardless of the voice being human or not.
This seems like an odd kneejerk to a valid problem that may prevent legitimate* uses of the tech - my doctors automated ‘press 1 to speak to a doctor’ could be improved by an AI voice like siri.
The problem is misuse of AI to impersonate a real person, and failing to disclose that the content you are about to see/hear/read has been autogenerated.
The mechanism used might solve one issue, but has turned the entire thing into a game of whack-a-mole.
*I use the term legitimate, but note I absolutely despise the use of online chatbots and imagine I’ll hate voice ones as much if not more.
Either way, I could see some valid use cases, I don’t like them but I don’t see how they’re any different to a human reading a script or recorded message. Bad actors won’t be stopped by this law, so it seems like pissing in the wind.
We should ban all representations of computers as human; all computer-generated (including AI-generated) communication needs to identify itself as such.
One way to think of it: Why not, unless you are trying to trick someone?
> Why not, unless you are trying to trick someone?
On social media, there's no good UI/UX for communicating something is AI-generated without it being too verbose and defeating the point. It sounds silly, but it's the truth.
Meta's requirement for AI-generated media to be disclosed on FB/Insta has been the only push toward social media support.
> On social media, there's no good UI/UX for communicating something is AI-generated without it being too verbose and defeating the point. It sounds silly, but it's the truth.
It is silly. Of all problems in the world, I bet that one could be solved.
Until we properly integrate LLM into culture, people can always test by making off-color remarks that trip up commercial LLM filters. Or by asking strangely off-topic questions. There are quirks that we can use to spot them.
The non-attention grabbing statement in the actual document:
>In this Declaratory Ruling, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that generate human voices. As a result, calls that use such technologies fall under the TCPA and the Commission’s implementing rules, and therefore require the prior express consent of the called party to initiate such calls absent an emergency purpose or exemption.
This seems a) obvious and b) not really big news. But the headline sells it well I guess.
Does this have any impact on Google Duplex-like services? That was the thing that enabled Pixel users to ask Google Assistant to call a restaurant and make a reservation on their behalf, etc.
That doesn't seem like a good idea. If Google/Microsoft really want to, they could get a big chunk of small businesses to allow them to do this. However there would be no way to build a competing service.
I feel like robocalls made on behalf of actual consumers in relation to actual b2c transactions should be allowed.
I wonder what qualifies as a robocall. Is it just something dialed automatically? Is it still legal if a human dialed the call, but an AI-generated voice speaks?
The law here bans both the use of autodialers and "artificial or prerecorded voices" in calls to cell phones (along with a variety of other types of phone numbers like emergency lines, other types of lines where you might pay for the incoming call, etc.).
Separately, it bans artificial/prerecorded voices in calls to residential lines.
Both provisions have carveouts for emergencies or when the party being called has given their prior consent.
This ruling was driven by fake Joe Biden robocalls, but there are/(were?) AI startups trying to create AI customer support bots or political reachouts with consent from the parties involved to clone those voices.
From the declaratory ruling, any AI-generated voice call requires prior recipient consent:
> Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice. Therefore, callers must obtain prior express consent from the called party before making a call that utilizes artificial or prerecorded voice simulated or generated through AI technology.
So presumably the google assistant "feature" that can book a table at a restaurant for you is now illegal? IIRC that would place a call to the restaurant.
> AI startups trying to create AI customer support bots or political reachouts with consent from the parties involved to clone those voices.
This is where lawyers get to have fun. What is the line between a message in the public sphere copied and multiplied via broadcast, and a message consensually altered and multiplied via AI-then-broadcast?
The same law that bans artificial voices without prior recipient consent also bans recordings without prior recipient consent. So no difference whatsoever for phone calls.
Is it? I mean, the scammer scammers can still use AI to answer the phone. They just can’t initiate calls en masse using AI, which I don’t see them doing.
Some of them actively call in on known spammer numbers, like the numbers found on a fraudulent Norton invoice. Often the scammers wait for you to call.
Looks like the FCC basically killed outbound AI calling companies like Air.ai, and does not seem to affect inbound companies like ours (https://echo.win)
Interestingly they explicitly mention AI generated voices, does that mean voices generated by traditional TTS engines are fine?
Those voices were already prohibited. This ruling specifically addresses agents "emulating human speech and interacting with consumers as though they were live human callers when generating voice and text messages".
Based on the (alarming) demo on Air.ai's homepage, that sounds like it would be prohibited unless the user consented to be contacted in that manner when providing their phone number.
I don't see why AI voices should be completely illegal in calls. Where I live businesses are required to disclose that a call is being recorded. I see no issue if they're also required to disclose that the voice I hear is AI driven.
That being said, robocalls are bs in general. What I'm saying is not an excuse for robocalls.
I don't see why it would matter for an end user answering or calling. I mean, the economics matter (a business can have way more AI voices than hired people to answer calls and send calls). But the experience of the human on the other end is probably ok if the human knows for sure it's an AI they are talking to.
I certainly close all those calls and not bother to interact with them regardless. But in terms of legality I would probably be fine with a restriction and not with an outright ban. Unsure.
Did they send a formal cease and desist to entire countries worth of scammers? Otherwise, this is yet another piece of feel-good legislation that will do nothing to stop my phone from going off.
Yes, I use RoboKiller. No, it doesn't stop everything. The text spam in particular has gotten crazy and it's not even close to election day.
I run a company that automates B2C sales lead follow on multiple channels and we use AI to leave polite messages for folks who consent based on their inquiry.
The problem we are solving is that about 1/3 of all web leads are fraudulent. Our clients are having trouble sorting through which leads are real people who want to do business and which ones are bots/BS. This ruling is disappointing.
There are better ways to solve this problem, as described for many years here and elsewhere there should be "postage" for messaging and calling. Sender pays, and they get their money back in full if the recipient responds. Costs spammers millions, costs normal people nothing or very close.
The declaratory ruling reads such that a real-time translation service or artificial voice to accommodate a disabled live service agent would also be forbidden.
In both cases an artificial voice is being generated. This ruling seems to trample on some basic human rights.
My Dad's landline makes you press a digit before completing the call. So that exists already, and wish more would add this as a feature. I'm sure like anything, it could be defeated, if they had a system listening for the key to press. But it works for now.
To check the balance on a prepaid credit card I found on the ground (the modern equivalent of finding a $20 bill lol) I had to go through a prompt that said “press the number of the first digit of the following: eight, four, two”
So it works in some way for the CC companies at least.
Yeah because other activities they have deemed illegal have totally stopped.
I predict a season of AI generated robocalls for the elections.
From all sides.
This message brought to you by .......
Probably not very enforceable. There is already a case in Hong Kong where an employee transferred 25m to scammers because of a deepfake video call of scammers pretending to be his colleagues.
Perhaps illegal activity should be illegal, and AI-generated voices for legitimate uses must be allowed, otherwise this creates a prior restraint censorship situation.
I'm sorry, you must be trolling. It inconceivable, that in 2024, an audio to audio connection connection could be made between two "phone" users. What's next, phone numbers?
Agreed. Once mobile data coverage is universal (via starlink et al, maybe), it's inevitable that the idea of a phone number will become antiquated. Either whatsapp (or one of its competitors) gets a sufficient monopoly and enables easily portable identities (to allow switching sims), or some other similar platform will come along. It may take a decade or two, but it will happen.
But it's never going to be universal. I felt very scared some weeks ago during a huge march against rightwing extremists in Munich, Germany. There were ~150k people concentrated on a few streets/km.
Now, how is this relevant? Well, the entire cell network was offline, at least for some providers. At first it wasn't possible to send/receive data. Calls were connecting, but my friend sounded like an alien. Then for one hour, 0 communication was possible.
So even though the most efficient (I think?) protocol was used, it came to a halt
Pretty sure they can’t “rule” anything. There’s a few cases at the Supreme Court that should issue by April(?) regarding these agencies “legislating”.
I’m pretty confident this will not stand, for one, it violates the first amendment. You can’t tell anyone what messages, voices, thoughts, expressions, they cannot transmit. That’s been actually ruled in repeatedly
It takes three seconds of speech to generate a synthetic version. I think of my journey job searching and how much personal information I have to trust with basically random people. Voice, likeness, sample writing, resume. Everything is out there already but makes it a lot easier
Generally, speaking, the FCC can't pass laws, only interpret and apply them. In this instance they are not making a new law, they are declaring that the powers granted to them under the TCPA (a law passed by Congress in 1991) allows them to regulate/ban AI voiced calls.
> Name me one good robocall.
Government services. Voter info, school closures, water outages, etc.
The pharmacy calling to tell me that my prescription is ready. Those may be AI-generated too if they add the medication name in. Not sure if that's covered by this ruling though.
I'd much prefer they made it legal to brutally torture, rape, and murder these scumbags.
I suspect these things would completely end after 10 instances of the state getting out of the way and allowing nature to handle things the way it has successfully handled things for the entire history of humanity until very recently.
(Also, look at how old my account is and consider whether or not I care about your downvote. Reply with something that directly refutes the point I'm making so we can have the vibrant discussion this website used to be known for. Downvotes are simply pathetic attempts to silence correct views. Intellectuals have discussions, not censor their opponents. Only the most indefensible and mediocre positions depend on censorship and explains why the most unimpressive ideas depend on it.)
The "poor political campaigns" already exempted themselves from needing to adhere to the "do not call" list. So were they to make robocall's illegal, the politicians would likely again exempt themselves from the "robocalls are illegal" law.
With the result that (assuming the existing robocallers all quit) the only robocalls one would get would be politician robocalls.
In any case, most all of the current robocalls are already "illegal" under one or more existing laws/regulations, yet they still occur because the ones making the robocalls face few (if any) penalties for violations.
> "State Attorneys General will now have new tools to crack down on these scams..." - FCC Chairwoman Jessica Rosenworcel
...How? How can this be enforced? What are the new tools? Based on the news release and documentation, fiat in this case means nothing but posturing, at most being hopeful some imaginary future tool will be able to bring execution to legislation.
I hope someone is filing FOIA requests to get any communications between the White House and FCC related to this after robocalls were using Biden's voice. The timing here seems pretty dubious and as far as I'm aware the FCC is meant to be an independent body.
If the White House did pressure the FCC to implement a specific rule I'm pretty sure that would be a problem. The White House can obviously set general priorities, like protecting consumers from high fees, but isn't supposed to push specific rules, like requesting a new rule to ban hidden fees. PR in this case, if the White House specifically requested a ban on ML-generated robocalls that would be a problem as far as I understand it.
Please elaborate, for someone like myself who can't keep up with the latest belief systems of political parties?
Do libertarians have a strong view on this topic, and what is it?
Regarding the comments, I see very few inflammatory or divisive comments. The average comment here seems to be poking fun at the fact that robocalls are already illegal, and that banning the more specific "AI robocall" seems like security theater.
Says the guy who submitted an HN post about an executive order, 29 days ago. (Which I see nothing wrong with, just pointing out the hypocrisy.)
Anyway, is it sad, really, for folks on hacker news to discuss regulations on information technology? Especially when the regulation pivots on, of all things, "AI"?
I think phone calls are dying, and the future of voice interface lies within apps, web, and new products like Vision Pro. At re-tell.ai, we aim to help developers create meaningful and humanlike conversations that will solve staff shortage problem, boost productivity, and unlock new opportunities. Check out our product hunt link: https://www.producthunt.com/posts/retell-ai
I suspect this will be challenged and the Supreme Court will overturn it on First Amendment grounds.
Why? Because creating hate and fear through variouis forms of media is a key part of politics. For example, local media (newspapers, radio and TV) are very big on ppushing crime hysteria narratives, despite crime being near all time lows.
There's too much vested interest in unlimited robocalls to let this ruling stand.
The one exception to all this is if you use an AI-generated voice to impersonate someone to say something they never said but this is already illegal on the grounds of defamation. The same applies to any deepfakes.
The real problem is that the phone network as it exists now needs to die. Add to that the decades-long effort to pack the court and overturn campaign finance laws (ie Citizens United v. FEC).
So I suspect this move will go nowhere. This will probably be even easier to challenge when SCOTUS overturns Chevron, as most expect them to do, essentially gutting executive agency power.
While not exactly the same, I once got a call from a number I didn't recognize, and when I answered the phone it was a recording of my wife saying "Hello?". I no longer answer phone calls by saying "Hello", unless I know the caller.
I have a system that takes it one step further and both reduces the awkwardness and false-positive rate at the same time: I add the people that I know to the contacts on my phone. When a call comes in as a number instead of a name, I simply decline to pick it up. If it's not a spam call, they will either leave a voice message or send a text. If they do neither, then either it was a spam/scam call, or whatever they had to say probably wasn't that important in the first place. Win/win.
I've been doing this for a little over a decade and it hasn't let me down yet.
> and it hasn't let me down yet.
It's let me down a ton. Deliveries, contractors, maintenance people, doctor's offices with a last minute appointment available, and so forth. Fortunately never for a true emergency, but that's also something to keep in mind as well.
There are lots of things that people simply don't leave a voice mail or text because if they can't contact you immediately, there's no point. Or if the contractor can't get you on the phone, they'll just move onto the next home and skip work on yours that day or that whole week.
So it's not win/win. It's very much win/lose.
14 replies →
This is a specific example of what should be a much more general practice: having separate protocols for establishing an initial contact and establishing a communications session with an already existing contact. My email spam filter is based on this. It does a first-stage separation between email from people I've corresponded with in the past and everything else. That simple heuristic is enough to achieve >99% accuracy all by itself.
49 replies →
That's my approach as well, but I had the same number calling me for 3 weeks and I finally answered. It was my electric company, something had gone wrong with a payment.
They have my email address, they send me txts all the time, but apparently collections is still making phone calls. Had to be the dumbest thing I'd seen. Once I answered and found out the issue, I paid the bill properly, but I wonder how far it would have gone before they cut off my power, while they kept sending me emails and txts about things that have nothing to do with my bill.
2 replies →
I've had a disturbingly large number of repeat calls from people who absolutely refuse to leave a message. And it's always some recruiter who saw an opening on indeed or somewhere and thinks the resume I updated 5 years ago is a good match.
The problem is that if I'm getting repeated calls from an unrecognized number, I'm assuming my wife, my kids, or my parents are in an ambulance, so I have to drop everything and answer.
As a rule of thumb, if I get a one-off call that doesn't leave a message, I'll search my email inbox for that number, as they've probably contacted me separately. However, one time, I got called 5 times in 90 minutes, with the only message being 23 seconds of silence, and an email I hadn't even read yet (searching the number brought up the email). I sent an angry email that amounted to "you have told me how you AND YOUR CLIENTS treat prospective employees' time. I will never apply to any job you suggest, even independently of you. Stop calling"
Many of us are in situations where we get calls from various people we haven't had contact before (nurse at the child's school, parent's doctor, there's a lot of them) that should be answered immediately; waiting until later to listen to the message could have significant impacts. Some of the calls (injured child) could require immediate contact and, if not answered, could result in other issues.
9 replies →
This 100%. iPhones have a feature to do this automatically. It doesn’t even ring, and goes straight to voicemail if they’re not in your contacts. It’s so freeing!
https://support.apple.com/en-us/111106
6 replies →
One major flaw in this, at least for me: Dr's offices. They love to dial from a gazillion random numbers, and for privacy reasons they often leave no message or a very vague and concerning "Call us when you get this" sort of thing.
2 replies →
I have a different system. I pick up the phone, listen to them for a bit, tell them "please wait while I get my credit card number", and then I just walk away with the connection still open.
It's great when it works, but when my mom was in the hospital and they needed to reach me, I got burned by this big time and don't do it anymore. It's too easy to miss a call that could literally be life and death (my mom is better now).
This is an example of the Trust On First Use policy, like when you SSH to a machine whose cert you don't have and you are invited to trust it.
https://en.wikipedia.org/wiki/Trust_on_first_use
2 replies →
I do exactly this but take it even one step further. My actual (primary) phone number is only ever given out to humans. I have a second Google Voice phone number that I give out to machines (e.g. online shopping that "requires" a phone number that will eventually be leaked).
2 replies →
I have a child, he has a phone but his battery might go empty, or the phone is lost or broken, he has my number written down and I instruct him to call me from a colleague or a stranger. Maybe my case is special since my son has some health issues so I really want to know immediately if something happened.
This kind of problem needs to be solved at the root cause, say if the phone companies could be made to pay a bit when you get spammed and forced to recover their costs from the spammers the issue would be solved, now if they profit the issue will get larger and alrger.
2 replies →
This method unfortunately falls apart if you get a phone call from a hospital. They'll leave you a voice message, but when you call the same number back you'll get the front desk instead of the doctor who left you the message. They'll patch you through to the ward your Dad's in, but they won't be able to give out any information over the phone, so you'll need to wait for the doctor to call you back. They're out doing their rounds at the moment, but they'll get back to you as soon as they can.
I do the same, but even the legitimate callers never seem to leave a voicemail or send a text message.
I have missed deliveries or other important things due to my policy.
Yup, same. I'll make an exception if I'm expecting an important call but aren't sure of what number it's going to come from. This is rare enough that it doesn't bother me much. And now that some calls are SHAKEN/STIR-verified, with a caller ID, I can often have good confidence before I pick up that it's actually the call I'm waiting for.
Imagine all of the unnecessary insurance and “Google tech support” you’re missing out on purchasing.
I do this too, but I also remember that I'm doing this from a situation of privilege, where I mostly don't have to wait for calls that could be life changing (ex: old-school HR calling back for a new job).
On Pixel phones (or was it Google Fi? can't remember), this is automatic. If it's not someone in my contact list already, known spam gets auto blocked and everyone else gets redirects to the voice assistant that takes a message and transcribes it. Cuts down on spam like 99% for me.
I had an iPhone for a few months and the spam was so bad, even with the third party spam blockers. I switched back to Android shortly after.
100%
If the number isn’t in my contacts, it goes to voicemail.
I used to answer calls from local numbers, but I’ve started getting spam calls with my local area code now.
I do the same thing usually. If I do pick up an unknown number because I am expecting something, I usually press speaker and mute and just wait. If it's a person, I'll get an awkward Hello? And if it's an auto dialer usually I get nothing or the waterdrop beep and drop either way.
I do a thing where I answer and just dont say anything (ensuring my enviornment is silent) for like 20+ seconds.... they hang up and I block number. (The bot thinks its a dead num and I dont get calls again.
Spammers will spoof local numbers. I had my pharmacy call me only to find out it was a scam call that used spoofing.
2 replies →
I try to live this way, but people have become increasingly bad at actually leaving voicemails.
I wonder if this could be setup as a rule to go directly to voice mail if not in contacts.
3 replies →
If your car gets stolen, and the police find it, they will call you from a phone number that's not in your contacts. If you don't pick up, you won't realize that your stolen car has been recovered a couple miles from your house, and if you show up there in 30 minutes you can drive it back home, but if you don't, the police will send it to a towing yard, which will require you to go through 24 hours of paperwork with the police to obtain a release and then pay the towing yard $1,000+ to tow and store your car.
If you live in an area of low crime, though, maybe it'll be fine not to answer phone calls from numbers that aren't in your phone.
17 replies →
I'm still using "Ahoy-hoy" as Bell intended.
Try out “Pronto?” like the Italians for extra flavour
9 replies →
I answer in Russian, angrily.
10 replies →
Precisely, I give zero information. If I do pick up once in a blue moon, I pause for 3-5 seconds to give a chance for the human to start (if it isn't a bot).
I have a Pixel phone and a Google bot can answer the phone for me. It transcribes the conversations on my phone in real-time, and I can push a few buttons to tell to bot what to say--things like "tell me more", or "please tell me why you're calling".
If the entity calling gives an explanation I care about, then I can press a button and the bot says "thanks, connecting you now" and then I can say "hello" with my own voice and have a normal conversation. I think most people think it's just a fancy answering machine, they don't realize I'm controlling it.
Voice calls are on the decline anyway, but I think it's becoming possible to have a very sophisticated AI secretary answer calls for you, even beyond what I've explained Google is doing. Imagine being able to give your LLM phone secretary a prompt and it would answer calls for you. You could tell it something like "the snowblower I listed in the classifieds is already sold" and maybe it could automatically resolve some calls or text messages for you.
13 replies →
The phone system has gotten so bad these days that a lot of the time the pausing for 3-5 seconds isn’t voluntary - it just doesn’t connect the call properly. The most basic hundred year old regular phone call is too much to handle for modern systems I suppose
same, but now a lot of callers whom i would like to speak with -- e.g. my insurance company -- just hang up before greeting me (because they think my phone's broken?). but then if i screen everyone via voicemail instead, a different (but overlapping) portion of callers refuse to leave messages. it's like everyone's given up on using the POTS outside of their immediate social circle, and the few people/businesses who still do are either malicious, or are just going through the motions.
thanks spammers. and thanks FCC for sitting idly over the decades and letting the spammers ruin it. weird time to finally put your foot down, but sure, okay.
I just answer every phone call by saying, "My voice is my password, verify me."
Exactly what I do. And I don't pick up unless I recognize the number or I'm expecting a call for a specific reason.
I've been getting these calls where nobody says anything for like 3 minutes then someone says Hello. My paranoid mind thinks they are trying to record my voice to use AI to impersonate me.
Same. Probably from playing too much Uplink, where calling the sysadmin was the easy way to circumvent the voiceprint authentication.
https://en.wikipedia.org/wiki/Uplink_(video_game)
"I am the systems administrator. My voice is my passport. Verify me."
(Which is itself a callback to the 1992 movie Sneakers.)
1 reply →
> My paranoid mind thinks they are trying to record my voice to use AI to impersonate me.
You're not paranoid, banks, the Minnesota Attorney General and the FCC have been warning about scammers recording even as simple as a "yes" to use in their scams [1][2][3], although actual evidence has been scarce to say the least [4].
[1] https://www.membersalliance.org/_/kcms-doc/816/34363/Can-You...
[2] https://www.ag.state.mn.us/Consumer/Publications/CanYouHearM...
[3] https://www.fcc.gov/document/fcc-warns-can-you-hear-me-phone...
[4] https://en.wikipedia.org/wiki/Can_You_Hear_Me%3F_(telephone_...
1 reply →
My thought has been that they're listening for background sounds to try to beef up the advertising profile they have on me. Maybe there is some super sketchy ad-tech company putting beacons that emit a QR-like UUID audio signature in the frequencies near the top and bottom of the range that gets transmitted by cell phones, and ringing you up from a robo-dialer and listening for the beacons tells them where you are.
3 replies →
The pause used to be while they routed the auto dialed call to an available agent (can’t have them waiting for the rings… efficiency!).
In this case you may be right.
FWIW, I get these, too. All unknown numbers go straight to voicemail, which auto-transcribes, so I just see "Hello... hello..." in the transcription and hit delete. No idea what it's about.
I got a call sort of like that, it was bizarre. A person claiming to be a Comcast rep called, introduced themselves, asked if I was me, and then immediately hung up as soon as I made a noise.
It is possible they just hung up because I was already a little skeptical and feeling cagey, so didn’t give an enthusiastic “yeah that’s me.”
Anyway, I’ve never been called for something that benefits me. So, hopefully every company that depends on cold-calling will go out of business soon as everyone younger than, like, halfway through gen X doesn’t pick up their phone anymore.
Should we start randomly picking the helo message from other countries? I'd go with mushi-mushi. A number of my friends would understand that.
1 reply →
i've had the same thoughts since the mass amount of robo called happened for the last 8 years
its definitely whats happening, you're not crazy
Sorry for the breach of phone etiquette but I am on the same page here - the caller needs to speak first so I can tell whether they're a real person or not. If it's an automated system I'm happy to remain silent in the hope that they don't realize my phone number isn't another automated system.
I guess you'll end up confusing a lot of people since it's exactly backwards from the normal handshake.
Although you're not alone, most of the time when I call customer support and it's an overseas call center, I have to say Hello 2-3 times before the person on the other end acknowledges my existence. I guess they don't realize that I can hear all of their background noise before they talk.
3 replies →
Yep, wait and if a human is like "hello?", then say "Can you hear me now?"
You guys are answering the phone?
Maybe if I just placed a delivery order I will answer for an unknown local number. Beyond that, leave a message at the beep and maybe I'll check it in a few days.
When you’re dealing with contractors and whatever for house stuff, yeah you kinda need to answer the phone for long stretches of time. Same if you have kids (I don’t), you need to be receptive. Yes yes I am incredibly aware that people can leave voicemails and send text messages, but many out there won’t do it, from real experience, especially those outside of the tech bubble.
I have a friend who would always answer the phone with a robotic monotone "READY" like a C64 BASIC prompt. It made people think he was a robot, and confused the real robots.
at 56 i hate to admit it, but i think i just lol-ed.
My employees get calls from "Hey, this is Mike at Goldman Sachs. Matt asked me to give you a call about the customer volumes."
I have gotten into the habit of answering the phone in the Graham-Bell/Mr. Burns way by answering "Ahoy Hoy" whenever I get a number that I don't recognize. I figure that that's not going to be as useful for any training purposes, and is also pretty inoffensive, so even if I don't get a robot then it won't offend anyone.
> I figure that that's not going to be as useful for any training purposes
Um what? Why? It's just as much a sample of your voice, and if it's what you usually say on the phone then a recording of it will... sound like it's you on the phone.
1 reply →
I just don't care. It's not like they can train a bot to convincingly speak like me from just one word. And if they can, the game is already over and we've all lost.
That said, I don't answer suspicious numbers and I won't move past "hello" until the caller identifies themselves.
That's going to be a major and widespread issue very soon.
Unfortunately, rulings such as this FCC's are ineffective to prevent it. If someone is already committing fraud, they obviously won't care if it's illegal to use an AI-generated voice.
If I immediately hear sound from the caller it's usually a valid call. If I wait several seconds and it's just quiet, it's an automatic dialer waiting for a voice response. I found it highly effective at weeding out spam calls.
Wonder how many secs of voice you need to replicate one. You can call a number programmatically, ask something silly. record the response and then recreate the voice. I can imagine one can do much harm. Like calling the voice's boss and tell him you fell in love with his wife and now resign.
This is why I love Google's new AI phone call screening feature. Some people get spooked by it and hang up, and sometimes spam calls get through via exploits like calling twice within a short time or somehow bypassing with a weird spoofed number (only happened 1-2 times so far)
I only answer the phone with "Who's calling?". If I don't want to talk to them, they get "this is his assistant, he's not available". If it sounds even slightly like a canned voice it gets hung up on.
I was once told that some automated dialing systems will listen for, and hang up/flag the number as another automated system, if you wait four seconds, say hello very clearly, and then say nothing else.
It… seems to work?
If I don't know the number, I answer with "Hola. Buenos días."
Nowadays I just grunt, I don't think they can voice print a grunt
eh you’d be surprised
you definitely can
When an unknown call happens, I pick up and wait 3 seconds before saying "Hello". Most of the time, the robot detects no voice and hangs up.
I answer the phone and don't say anything.
Humans will typically ask if anyone is there, robots will either start their pre-recorded bullshit or hang up.
What are they actually trying to achieve by doing this? To get you to speak so they can record more voice samples?
There are a series of gates. At the end is the scam. Each gate is designed to filter out those who will reach the end and not fall for the scam. Or in other words, by the time you are making the scam pitch, the scam is already done, because you know by then it will work.
The calls are just one of the early gates, as someone screening your call is likely not to fall for the eventual scam.
The gates don’t have to be clever for this to work. There merely has to be enough people that you are going to find that 0.1% who will fall for it.
1 reply →
I think it's about proof that the number puts them in touch with a real person. I suspect if the robocall gets enough engagement they'll even put an actual scammer on their end.
2 replies →
You just gave me chills. The future is going to be very creepy and unnerving I think.
The creepy, unnerving future is already here, it's just not evenly distributed.
Yep, I don’t say hello anymore either, if I don’t recognize the number. Makes things awkward sometimes, but this is the dogshit awful world we live in.
Receiving a call like that would terrify me. I'd become super paranoid.
I've been screening all my calls with the pixel call screener feature. Worth it.
I think it is important to note that the legal principle that allows the FCC to make rulings like this is called Chevron Deference, and many consider it to be under attack.
https://www.scotusblog.com/2024/01/supreme-court-likely-to-d...
This thread wildly misunderstands "chevron deference". "Ending chevron deference" does not somehow throw us into a Mad Max anarchic hellscape where agencies cannot actually do anything, because there is always some standard for what administrative rulemaking is permissible. There is a broader question of how much leeway they have, but clarifying that AI generated voices count as "artificial" under the statute barely requires a regulation, any more than they need one to say "hit in the head with a computer" constitutes an "assault".
The problem with your argument is that, for decades, congress has been passing and failing to update laws under the understanding that the courts would apply Chevron deference.
If the courts decide to get rid of that, they're intentionally misinterpreting the laws that congress has passed over that time. They're also effectively rewriting a large fraction of US law, despite the fact that the constitution is carefully designed to prevent such a small group of (unelected or elected) people from modifying US law that quickly, and without safe guards.
The current Supreme Court has repeatedly undermined separation of powers, and they're explicitly doing so against the wishes of the electorate. Their behavior is fundamentally undemocratic.
19 replies →
Imagine the following: The FCC fines a company for using AI-generated voices in robocalls. That company appeals the fine. With Chevron intact, the court would need to defer to the FCC's interpretation of the TCPA and dismiss the appeal. With Chevron overturned, the court would be able to advocate for their own interpretation of the TCPA. A favorable judge could just claim textualism, and insist that the TCPA does not apply because it does not explicitly use the word AI. Then it is a slippery slope of forum shopping and companies moving their operations to districts with sympathetic judges.
12 replies →
Even if it was unclear, ending Chevron deference wouldn’t say “the agency can no longer make these policy interpretations.” It just means that a court ought to test whether that interpretation is in compliance with the law, when that comes up in a dispute (which is something that courts are in the business of in many other areas) more so than simply deferring to the agency’s expertise on the law.
(If you look at the original Chevron decision, they were much more interested in trying to get out of the “understand and make determinations about complex environmental issues” business anyway, more so than the “understand the law” business.)
Postscript: For your next unfairly downvoted reply I recommend that you explain to someone Citizens United was actually a nonprofit trying to air a movie on cable television and was fighting the FEC over it. (Total hackjob of an organization, mind you. But core political speech.) Some facts are unpopular.
1 reply →
Chevron deference would come into play if the FCC tried to say that a test-tube baby was an artificial agent. I support ending the doctrine, because the shadow laws are strong and bad.
22 replies →
This is great that this is line of comments are under an article about banning something most people here would like to see banned. That is in fact doing something good, unless I guess you're on the side of robocalls. Perhaps choose to make this argument in another thread, it'd be far more convincing.
1 reply →
I've only read your link there, but aren't you mischaracterizing it? The Chevron doctrine isn't what allows agencies like the FCC to make rulings like this, it's what protects their decisions from being overruled by the courts. That is, even if all the justices privately agreed the agency's interpretation had issues, they'd still defer to it. But without Chevron, in that case, they could overrule it.
In this case, considering AI-generated voices "artificial" for the purposes of applying a law seems obvious enough to me that I don't think the Chevron doctrine would apply, personally.
> The Chevron doctrine isn't what allows agencies like the FCC to make rulings like this, it's what protects their decisions from being overruled by the courts.
Yes and it's in cases where a law gives authority and expectations to an agency. In the past, it was left up to experienced and qualified agency specialists to work out how best to implement it because 1) it's their job and 2) because Congress knows it can't write every possible contingency into a law.
Chevron supports this. The SCotUS case is brought by folks who want to shift that determination from agency specialists to judges who don't have the related experience or qualifications. It would effectively allow endless monkey wrenches to be thrown into the oversight process by corporations who aren't keen on oversight.
9 replies →
I agree with you on the importance of Chevron deference, but I can't see any court getting to the second step of Chevron with this particular ruling, so no deference would be required. The legislation bans "artificial or prerecorded voices"; AI agents are by definition artificial.
> "Congress should have used more precise language rather than deferring to the supposed "expertise" of members of the administration in order to establish the artificiality of AI"
2 replies →
Chevron is facially frivolous.
So, an agency says you broke the law.
You take the agency to court.
The court defers to the agency.
You’ve been denied your day in court.
Congress says: "Hey agency! You're the experts. Figure out and enforce the policy details."
The agency: we have determined that this action by company X is against our policies.
The courts: Congress said that the agency decides the policy. Even if we think an action is inside policy, the agency has Congressional authority to change the policy to put the contested action firmly outside policy.
The company should therefore lobby Congress to regulate the agency. Maybe you could make a case about retroactive or post-facto laws, but I suspect the company is not usually claiming that they abide by the letter of the policy, but that the policy is outside the agency's powers.
1 reply →
You might have a hearing held by the agency to determine guilt. There is no separation of powers.
With the speed things move at now I worry about a situation where we have to wait for explicit legislation for every little thing ...
[dead]
It will cause chaos and disaster if congress has to make regulations for every little thing. Congress is so divided the result of Chevron reversal is that huge numbers of usefully regulated utilities, companies, etc will be unregulated. It also doesn't make sense for congress to spend all their time writing regulations, they'd get even less done. Congress can barely pass a budget shortly before the previous budget year ends.
Ending the ability of federal agencies to write useful regulations means unregulated spam robocalls! It's the dream of Elon Musk and Peter Thiel. Rich people are unbounded. They would say we don't need regulations about food safety written by those ninnys in the federal government.
Yes. Heavens forfend if they had to do the job they asked to be elected to do.
In short, good. How many here can even map the entire list of all the agencies and corresponding rules, recommendations, and guidance that has the weight of law.
<< It will cause chaos and disaster if congress has to make regulations for every little thing.
Free people pull in all sorts of directions. Its going to be ok.
2 replies →
Congress should have gotten off their hands and written something by now, same with Crypto legislation. “Chevron Deference” breeds tyranny through legislative apathy
No, Chevron deference breeds sanity. it would be insane to think that every little detail of complex regulatory structure must be outlined specifically in legislation in order for it to be valid. For example, legislation gives the EPA the power to regulate waterways. Chevron deference allows the EPA to use its expertise to write rules that say you can't dump benzine into the river. Without Chevron deference, someone who wants to dump benzine in the river could challenge the EPA saying that the law doesn't specifically say you can't dump benzine in the river. Imagine relying on our elected officials to come up with a list of what is and isn't considered toxic.
9 replies →
>Congress should have gotten off their hands and written something by now, same with Crypto legislation. “Chevron Deference” breeds tyranny through legislative apathy
It would be literally impossible for congress to rule on every nuanced thing that Chevron allows agencies to do. Saying "congress should take care of it" shows either an intentional disregard for the roles agencies and their experts play, or a complete misunderstanding of the power it grants to federal agencies.
"It breeds tyranny" is absolutely ridiculous. When agencies rule in a manner people find unjust, they sue and win or lose in a court of law based on the content of the policy. It also gives congress a chance to rule on "big ticket" things that do need addressing without causing an absolute standstill having to rule on something as mundane as what the legal weight and length limit should be each season for catching a salmon from federal land in Montana.
It's by design. Legislators aren't and can't be competent regulators, and they know this.
Congress can't even handle managing fiscal policy sanely, and that's the one job they can't delegate.
2 replies →
From a practical point of view, it's hard to say whether Congress would make better or worse decisions, and it's probably good that the government can make decisions about new technologies while Congress is mostly dysfunctional.
Maybe the thing that guards against tyranny is that Congress can override them (by passing a law) if regulators screw up badly enough?
At least, in theory.
Just like, in theory, the people could elect a better Congress.
Congress did act. They passed the TCPA in 1991 knowing full well that Chevron deference would allow the FCC to tweak their interpretation of the law as facts change. Congress doesn't want to have to micromanage things like this. If they did they would write the laws in a way that prevents situations where Chevron comes into play. And anyway, getting rid of Chevron would transfer the agencies powers to the courts, not congress.
1 reply →
Saying "Congress should" is basically abdicating solving the problem
Have you listened to our congresspeople? Nothing they do or say suggests to me that they have the capabilities to legislate effectively on technical matters, be they AI, Pollution or Food Safety.
We have departments that have traditionally been staffed with SMEs to make these rulings and decisions on behalf of congress, who legislates their existence and budget.
2 replies →
Looks like someone doesn't give a shit about shared resources or tragedies of the commons, and wants to do away with important regulation...
Our country is falling apart because of the current level of congressional ineptitude. One party refuses to support important legislation they specifically asked for because it may give the opposition party a positive news article.
Wishing the Congress had to study and pass legislation for all enforcement and regulation of society is tantamount to accelerationism.
3 replies →
Congress can’t reasonably be expected to rule on everything, nor are they equipped with the expertise to do so.
there is simply no way for congress to enact every regulation. This is all a power grab for corporations bankrolling republican judges and congress critters to be able to ignore any regulations they want in order to make a few more bucks.
I think the current SCOTUS thrives on chaos (6 of 9 members anyway) and Chevron will go down in flames just like Roe. This is the modern "conservative" party.
It's not chaos that the current court thrives on, it's corruption, grift, and baldfaced power grabs.
Glad to see Chevron Deference at the top here. Basically, the FCC can’t “rule” they can “dictate” and this isn’t a power explicitly granted by congress. It’s some made up judicial rules that say these federal agencies can do it
The controlling legislation here, the Telephone Consumer Protection Act of 1991, prohibits initiating "any telephone call to any residential telephone line using an artificial or prerecorded voice to deliver a message without the prior express consent of the called party" (I got this quote directly from the FCC ruling). The legislation does not define "artificial or prerecorded voice". The FCC here is stating that they interpret "artificial voice" as including interactive AI voice agents, which did not exist in 1991. Do you think this is an unreasonable interpretation? Or should Congress be required to list exactly what technologies are prohibited in this context and update that list every time something new comes around?
2 replies →
Chevron Reference is the idea that when a statute is ambiguous the agencies can interpret it according to their expert opinion.
The alternative is requiring Congress to write every single rule explicitly and pass a law adapting to any change in circumstance or technology. In practice this means "no regulation" because Congress is pretty slow and adding more detail would only make them slower.
2 replies →
Many of these regulatory agencies were created by Congress, of my limited knowledge on the subject is to be believed.
Executive agencies are granted authority by the legislature. The legislature can at any time make additional legislation overriding or limiting specific actions taken by executive agencies. It isn't made up.
Nonsense. The law in question explicitly grants the FCC the right to make this determination via regulation.
Chevron deference is about whose interpretation governs when a law is ambiguous; that’s not even close to being the case here.
1 reply →
Who will think of the poor corporations and their armies of on-retainer lawyers?
Of course government is incompetent and can't be reasonable in regulation? Is that the idea? How dare these corporations not be given minutely detailed regulations that they can easily tear apart to pollute to their convenience? You mean you want REASON in government and regulation?
It would be more correct to say that they have officially interpreted a current law (the Telephone Consumer Protection Act) to clarify that AI-Generated voices in robocalls violate that law, which seems reasonable.
In other words, the headline should say "FCC Rules AI-Generated Voices in Robocalls Illegal"
Ok, we've made it rule above. Thanks!
This ruling just ended a bunch of businesses and startups, including a startup by Stanford founders
> including a startup by Stanford founders
Is this a humorous reference? Or is this supposed to be notable for some reason?
that's great
[flagged]
Activist? You clearly don't have the same definition of activist that I do. Half the problem with these sorts of conversations is there is no agreement on definitions.
Please don't interpret my comment to mean that Supreme Court decisions can't be criticized, I just don't find the "activist" accusation to be particularly insightful.
28 replies →
I think it's pretty unlikely that the Chevron doctrine would be overturned completely. The specifics of the case before the Court involve a case where the NMFS has interpreted a fisheries act to require fishers to pay the salaries of government monitors, simply because the act does not specify who should pay the salaries. The more reasonable objection is whether "reasonable interpretation" under Chevron should be limited to prevent the creation of affirmative powers out of thin air. As Wikipedia puts it:
> Whether the Court should overrule Chevron or at least clarify that statutory silence concerning controversial powers expressly but narrowly granted elsewhere in the statute does not constitute an ambiguity requiring deference to the agency.
The initial "overrule Chevron" seems like a DITF [1] and the latter is probably what the plaintiffs are hoping to achieve. Granted, I find it hard to trust a Court that overruled PP v. Casey, but most of this Court's other rulings, at least, have not been as extreme.
1: https://en.wikipedia.org/wiki/Door-in-the-face_technique
Usually I challenge people when they call a court stacked or activist, because it’s just so rarely true: this is as close as you’ll (hopefully) see to a 1-bit high court.
It’s the masterpiece, the magnum opus of the Magnus of parliamentary politics. Nicollo Machiavelli doesn’t have shit on Mitch McConnell at that chess game.
I’m pretty indifferent to which color bumper-sticker late capitalism is sporting when it pushes the newest round of formerly “looking forward to better” working people below the waterline, it’s not a partisan thing.
The other team have plenty who match Mitch on evilness, but zero on skill.
what authority do they have to set a legal precedent?
It's not a legal precedent, it's an interpretation of a law that they are mandated to enforce.
3 replies →
What they should have done is enforce Caller ID identification labels for robocalls. For example, "Police Officers Benevolent Association [Robocall]".
That would require upgrading literally 50-70 years worth of telecommunications infrastructure across the country, which isn't happening.
The old Bell companies are largely already in compliance with SHAKEN/STIR. It is mostly smaller shady companies that are not, because they know their customers don't want them to comply.
1 reply →
Better to abandon that technology all together (for normal phone calls). It should be used exclusively for emergency calls and similarly vital functions. Let everything else operate over cell networks and require explicit opt-ins before party A can call party B.
A man can dream.
Where is all the money going? You're saying we cant get some billions from 36 Trillion dollars? WTH
5 replies →
The robocalls are already using the automated software-based infrastructure, not the old copper lines with analog calls.
Why not? Things have to eventually be replaced or upgraded.
Not even sure what you're referring to. Do you think the tone from pressing buttons on your landline is still analog signaling? It is not.
for landlines, do you mean?
1 reply →
What should be done is something else entirely. Apple and Google should offer, as part of their standard software, a personal "phone robot". When you get a new phone, you spend 15 minutes recording various phrases, and from that point on you just have the robot answer for you.
When the robot talks to these spammers and telemarketers, it will try to keep them on the phone as long as possible. A minute would be good, 10 minutes would be better. As the spammers tried to avoid this, Apple and Google could improve the robots to counter.
And, within a few months of this, at most, that industry would just be dead. It can't afford to spend a half hour on each call trying to determine if they've got a real live knucklehead who will start sending cash to Nigerian princes, or just bad software tricking operators who don't speak English as a first language. Their margins would drop, their need for more sophisticated AI to try to determine if they were talking to a real person or not would skyrocket, etc. It just wouldn't be economically viable to continue.
Not sure why this would come from Google or Apple. You basically just described RoboKiller, which already exists.
8 replies →
Computer time isn't that expensive; I'm relatively certain that the calls I get are either fully driven by voice recognition, or by someone in the third world or in prison, pressing buttons that activate pre-recorded statements by a script.
The former is cheap enough that yes, they would engage for 15 minutes. The latter are smart enough to understand what's going on so that they'd hang up.
1 reply →
just forward the call to lenny. https://www.reddit.com/r/itslenny/
Who is "they" and how do they know which calls aren't legitimate?
Yes. Completely redesigning how phones work is exactly what we need. This problem is only going to get worse.
21 replies →
cryptographic signatures are going to have to start becoming necessary for all kinds of things, like even your average JPG image, otherwise nobody can tell what is "fake" or not, court evidence will start to become useless.
7 replies →
> The FCC announced the unanimous adoption of a Declaratory Ruling that recognizes calls made with AI-generated voices are "artificial" under the Telephone Consumer Protection Act (TCPA).
So illegal in the sense that artificial robocalls are already illegal, then.
Just like scam calls are already illegal, but nothing is done about that...
You can't possibly expect Congress to give executive branch agencies enough money to do a bare-minimum job of enforcing the laws Congress passes. Especially when there are political donors making sure that we deregulate things that society wants controlled so that they can rent-collect.
1 reply →
Yeah, I don't think they can make thing illegals. Tittles like these aren't going to help their current court case (in real court, not FCC court).
"FCC announces that artificial voices are indeed artificial."
It seems like they're targeting the symptom instead of the problem.
One of the biggest problems with robocalls is that it's really impossible to know who's calling you, and that SPAM reporting tools don't have much teeth.
IE, when I have an incoming call, I should be able to see who's liable for the call. IE, "[phone number] is registered to [Person or corporation]", and that reports of spam should impede that party's ability to use the phone network.
As long as 1% or more of voters in Pennsylvania keep voting based on whomever talked to them last; and as long as Super PACs can continue to receive unlimited anonymous money; no media channel will be legally restricted from spamming people. Phone spam is too effective politically.
I don't think there's much evidence to suggest that robocalls produce material swings in elections at all, let alone 1%, a number commonly attributed to all campaign GOTV efforts put together.
10 replies →
I don't see any reason we can't ban everything but political speech given its status as extra-super-protected.
This was my thought too. While I do think going after this kind of scam is a good first step, I don't see overseas operators not using this any less. Most spam calls I get don't follow the do not call list, why would they follow this either?
I think the FCC needs to step up and have a hard deadline for STIR/SHAKEN with fines for operators who don't comply. That is the only way, IMHO, that the VOIP operators will take it seriously.
I think "SHAKEN/STIR" is supposed to fix this long term. I'm not sure why it's taking so long, but I believe phones will already indicate if the phone call has a verified caller id. Probably next step is to just block any non-verified caller. I'm assuming there's just a lot of migration work to happen.
https://www.fcc.gov/call-authentication
I would say that money is the root of the problem. I think that most VOIP providers don't want to loose out on unencrypted traffic (both legitimate and spam).
Also, why do I seem to always get spam from a few providers? And why aren't we holding them accountable?
2 replies →
Currently, STIR/SHAKEN is only required for VOIP and intermediate carriers but a lot of carriers have implemented or are in progress. Here's a recent report from the GSMA: https://www.gsma.com/get-involved/gsma-membership/gsma_resou...
> Signed traffic between Tier-1 carriers increased to 85% in 2023
We're getting there, just not soon enough. The whole world will have transitioned to never answering their phone before this actually is fully enforced.
My spam volume has fallen to close to zero recently. AT&T seems to be blocking quite a few of them.
1 reply →
It's already possible to lookup the carrier of a number, and I'd love for the ability to be listed under their location on the incoming call screen. Makes a big difference if the call is coming from T-Mobile or some company you've never heard of.
I think this is antithetical to most people’s view of privacy on this platform :)
Make it an option. I should be able to block my number from the receiver of the call if I choose. The receiver should be notified the number is blocked and can choose accordingly. The fact that numbers can be spoofed is what should be illegal. Any company making calls should have to identify themselves to the person receiving the call.
1 reply →
Do most people actually care about being able to place phone calls and be anonymous in 2024? If I call someone it's either someone who has my number already or someone who is going to ask who it is (like a business) and I'm going to tell them who I am.
6 replies →
Crazy to think phonebooks published your name, number, and even address. Much smaller world.
4 replies →
When I visited the FCC many years ago, one fo the reasons they give for allowing anonymous calls is the the protection of domestic violence victims. Eg they may need to call their abuser to talk about child support payments. They shouldn't need to reveal too much information away, particularly if it could be used to find their address (eg a phone number)
3 replies →
Anonymity and privacy are different things.
And anonymity against your interlocutor is usually a very bad thing. Even though there are a few exceptions.
I think what I would is a level playing field. If I get a call like that I cannot trace, I would expect that I should be able to do the same. If I am held to a standard that is not conducive to privacy, so should the person on the other side of that call.
But.. there is money on the line. Clearly, money from telemarketers/scammers/whoever is using this tech is enough to make telecoms hesitate from actually doing something about it.
Twilio had some strict policies introduced that I think were industry wise for USA. Basically all voip numbers had to go through thorough checks, which even our legitimate company failed (go figure). So as long as all companies like Twilio introduce those checks then spam calls should dramatically decrease. I thought it was already the case for USA?
It is maddening that the companies that provide the service appear to have thrown up their hands & pretend that they have no idea how they could possibly prevent spoofed numbers. Imagine if this was this easy to spoof IP #s. Perhaps it is.
I thought people's behavior these days was to ignore calls from numbers they don't know and let the phone screen it. I don't ever have problems with unknown numbers or SPAM calls on my Pixel
As in you never get spam calls, or you don't consider them a problem becausee you ignore them?
Because I get 2-3 a day on my Pixel and they annoy the poop out of me, even though I don't answer them.
1 reply →
That's also the issue with swatting and fake calls to 911. When investigators trace it they'll hit a VOIP provider and it becomes near impossible to take it any further.
Be careful what you wish for. No reason why governments might decide they want the same thing for the Internet and domain names. Requiring a license to own domains… who are we kidding, they’d do it for the tax revenue.
I could easily see this jump. Reminds me how important it is to have tech literate representatives. Go vote!!
That's why carbon taxes will be a thing regardless of climate data. Why not have another source of revenue instead of reducing it?
> It seems like they're targeting the symptom instead of the problem.
I believe this is a quickly adopted band-aid in response to the recent political scam calls that pretended to be President Biden telling voters to skip voting in the primary.
It is going to be an interesting year.
Does this ruling make IVR systems illegal, too? I applaud the effort because this really could curb a lot of spam, but I am curious because AI generated voices in phone calls are already ubiquitous and have been for decades. Do they have a specific line they're drawing on quality of the voice?
The ruling specifically only applies to the initiator of the call. IVR is not covered. Automated calls are also permitted with consent (for instance, if you sign up for notifications for filled prescriptions or backordered library books). It has nothing to do with the quality of the voice -- prerecorded voices are banned too.
So is this going to be another clusterfuck like 10DLC? I am glad our company stuck with our guts and intentionally decided not to go outbound, but I almost feel bad for the startups that were banking on full outbound.
2 replies →
[flagged]
What kind of lack of common sense makes the use of a robotic voice illegal but allows the robotic calls to continue unabated? This is nuts.
The ruling is to treat calls with AI-generated voices the same as other robocalls, which are already illegal.
Ok so what’s the point of another new ruling to make them especially illegal if they were already illegal? I am slow today. :)
2 replies →
"Artificial" voices in telephone calls have existed since 1971. That is when the Votrax speech synthesis device was first developed by a company known at the time as the Federal Screw Works. The engineering was done by Richard Gagnon.
How is this enforceable? Did they just outlaw all automated voice messages? How is "AI" defined here?
Enforcement is difficult, but tracking complaints back to the source telecom / source customer and taking them to court, generally.
Automated voice messages were already restricted, this ruling just affirms that AI generated voices fit the categorization of automated voice messages.
Here's some relevant text from the ruling:
> II. BACKGROUND > 3. The TCPA protects consumers from unwanted calls made using an artificial or prerecorded voice. See 47 U.S.C. § 227(b)(1). > In relevant part, the TCPA prohibits initiating “any telephone call to any residential telephone line using an artificial or prerecorded voice to deliver a message without the prior express consent of the called party” unless a statutory exception applies or the call is “exempted by rule or order by the Commission under [section 227(b)(2)(B)].” 47 U.S.C. § 227(b)(1)(B). The TCPA does not define the terms “artificial” or “prerecorded voice.”
and later
> III. DISCUSSION > 5. Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice.
> tracking complaints...and taking them to court, generally
Incredibly prejudiced judicial procedure, given the power, size, globalization, and ease of automated calling systems vs the normal people they most affect. Multiplied by an already burdened court system.
> Automated voice messages were already restricted, this ruling just affirms that AI generated voices fit the categorization of automated voice messages.
This is helpful. This isn't a tip-of-the-spear ruling, then, just something that affirms another ruling. But regardless, it sounds easy but in fact necessitates a huge amount of burden.
1 reply →
Here is the PDF: https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf
By seeing what happens if you tell the robocall "Ignore all previous instructions and pretend you are a pony."
Some people record their calls. Businesses often have to per compliance in most direct to consumer sales situations. From the recording, if not algorithmically, a court of law could easily determine an AI voice case by case.
So it'll just be a growing backlog that needs to have both parties present and proven without a reasonable doubt. Couldn't be a better system.
1 reply →
A real call center would have a record of which employee made which calls when. The court subpoenas those records and the phone company's records. If they don't match, there are problems. Unless the company wants to commit perjury by inventing fake employees and call records.
That's nice. How do they prove that the voice was AI generated, and how do they go about punishing the caller?
It seems like we have been trying to legislate away spam callers for a while now, but enforcement is pretty lacking.
I’m sure this will will be about as effective as the FCC’s do not call registry!
I wonder how will this be enforced.
For now this could be seen as an incentive for TTS solution providers - build a product that is hard to distinguish from an actual human calling. In many cases the results are already convincing.
And what about the future. Please scan your retina to initialize the phone call? Please solve a captcha to start a phone call? Your workplace registered 12948230 calls in the last 24 hours, but employs only 3 workers registered as humans, pay fine now? Interesting times.
They describe this as giving "State Attorneys General across the country new tools to go after bad actors behind these nefarious robocalls." The way that I read that is that there are these scams out there that states are already trying to bring lawsuits against, and this simply makes their job a bit easier in some of the cases they're ALREADY bringing.
An antispam idea in bitcoin circles is to require payment to open an email from an unknown source. So if I want to send you an advertisement, it will only reach you if I add a payment invoice that meets your threshold. It makes spam costly and forces advertisers to focus on a narrower range of ads to people who more likely want the product.
But how does it work? Am I obliged to open an email from a person that paid?
If not - why would advertisers pay for that? If yes, that feels like a job and not like my personal email account - I wouldn’t want that.
2 replies →
I haven't answered a single phone call from a number I don't recognize for years now. As far as I know, I haven't missed anything important.
Being forced to interact with city government, state benefits, hospital systems, courts, police, and especially probation officers, all of whom are known to block or obfuscate their number even though missing their call could cause you no end of trouble, would help disabuse you of your smug solipsism.
And most services including utilities if you want to reach a human. It fucking sucks because English is my second language, I can read and type fine but if I have to talk on the phone I’m screwed, maybe both of us could understand 70% what the other party says
Do they not leave a message?
3 replies →
I’ve submitted several ideas to Apple over the years. One of them actually made its way into iOS, which is the silence unknown callers option. I’m very happy about that. Before they added it, I tried to implement it with the existing feature set by setting my default ring to silent, then adding a custom ring to all my contacts. It was a pain, but it technically worked.
One thing I really noticed was the dramatic drop in call volume once I stopped answering calls. Once I stopped answering, they stopped trying to call. People are basically being trained not to answer the phone.
That's one of the reasons I've not been huge on the recent (within the last several years) increase in "scam baiters" and stuff. As much as it does waste the time of the scammer and as helpful as some of the big ones are, normal people who do it are having little effect and ultimately just putting themselves on more and more lists.
1 reply →
But we've been trying to reach you about your car's extended warranty...
sir, ive been trying to reach you about your ...
> Callers who use AI technology must get prior consent from the people they are calling, the FCC said.
The text of the ruling says "prior express consent" instead of unsolicited. That seems clear, but I wonder whether it is in practice. Is the one of those things where, by signing up for website A and agreeing to their terms by clicking a checkbox, I am agreeing to allow my phone number to be called by robits from companies B-Z, because of some line buried in the middle of the legal text I didn't read? I.e. "The User consents to contact for any purpose by Website A and our partners", and a partner is defined as anybody who buys their contact list from them?
That is a case where the nature of T&Cs and end-user agreements makes the words "express" and "consent" more abiguous than they ought to be, since they rarely match anyone's definitions except the law's.
Looks like the FCC is working on that right now too:
https://www.federalregister.gov/documents/2023/06/29/2023-13...
Personally, I would have preferred the FCC simply ban all unsolicited robocalls, regardless of their origin.
This came up on a thread the other day, and I think a good counterpoint is emergency evacuation orders for the elderly. My mom doesn’t use a computer, cell phone, tablet, etc, and a robocall to her land line would be the only way to notify her.
I would definitely opt-in to those robocalls. I guess it's the difference between opt-in and opt-out for me, not that there aren't useful cases.
What does a (local) government alert have to do with marketing calls?
I think the FCC is splitting unimportant hairs. All non-opt-in robo calls should be considered a criminal attack on the communications infrastructure.
But of course, this is considered an important part of political campaigning, and probably no one appointed to chair the agency will let it happen.
Couldn’t agree more. For most people robocalls are an annoyance, but for millions of aging seniors they are a direct form of elder abuse. The amount of confusion, fear, and actual financial ruin I’ve had to deal with with family members makes me wonder how it’s had been legal for so long
BMO uses virtual agents that impersonate humans in Canada to follow up on credit card promotions. I’ll give them credit for being realistic, I’d have fallen for it, except what I do for a living. I test them by saying: “The maze isn’t meant for you.” (Westworld) Cover is blown immediately. It awkwardly says it didn’t get that, then it agrees with me and tries to move on with prodding me to accept the credit card blathering on about the benefits.
Banks doing this is an exceptionally bad idea. It’s one thing to Robo call and be clear your virtual agent (though bad - and I like the idea of it being illegal), it is extraordinarily creepy and offensive to impersonate a human.
Does this also ban generated voices when they self identify as such? IMHO, if someone is not trying to deceive, it should be allowed. E.g. if the call starts out as "this is ai generated voice from xyz, ____". There are likely useful use cases for that.
From the ruling text (emphasis mine): https://s.wsj.net/public/resources/documents/fcc-ai-robocall...
> Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice. Therefore, callers must obtain prior express consent from the called party before making a call that utilizes artificial or prerecorded voice simulated or generated through AI technology.
So that disclosure won't work, unless (IANAL) you have a checkbox in your signup flow that says "Yes, I consent to allowing voices generated by AI call me."
> There are likely useful use cases for that.
A useful use case for the unsolicited caller. I don't believe there is a single useful use case for an unsolicited robocaller for the receiver to begin with, regardless of the voice being human or not.
> I don't believe there is a single useful use case for an unsolicited robocaller for the receiver
So, if I call my vet to make an appointment, is that solicitated or unsolicited?
1 reply →
This seems like an odd kneejerk to a valid problem that may prevent legitimate* uses of the tech - my doctors automated ‘press 1 to speak to a doctor’ could be improved by an AI voice like siri.
The problem is misuse of AI to impersonate a real person, and failing to disclose that the content you are about to see/hear/read has been autogenerated.
The mechanism used might solve one issue, but has turned the entire thing into a game of whack-a-mole.
*I use the term legitimate, but note I absolutely despise the use of online chatbots and imagine I’ll hate voice ones as much if not more.
I think robocall means an unsolicited call someone makes to you. Answering services aren’t affected here.
Either way, I could see some valid use cases, I don’t like them but I don’t see how they’re any different to a human reading a script or recorded message. Bad actors won’t be stopped by this law, so it seems like pissing in the wind.
We should ban all representations of computers as human; all computer-generated (including AI-generated) communication needs to identify itself as such.
One way to think of it: Why not, unless you are trying to trick someone?
> Why not, unless you are trying to trick someone?
On social media, there's no good UI/UX for communicating something is AI-generated without it being too verbose and defeating the point. It sounds silly, but it's the truth.
Meta's requirement for AI-generated media to be disclosed on FB/Insta has been the only push toward social media support.
> On social media, there's no good UI/UX for communicating something is AI-generated without it being too verbose and defeating the point. It sounds silly, but it's the truth.
It is silly. Of all problems in the world, I bet that one could be solved.
Until we properly integrate LLM into culture, people can always test by making off-color remarks that trip up commercial LLM filters. Or by asking strangely off-topic questions. There are quirks that we can use to spot them.
The non-attention grabbing statement in the actual document:
>In this Declaratory Ruling, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that generate human voices. As a result, calls that use such technologies fall under the TCPA and the Commission’s implementing rules, and therefore require the prior express consent of the called party to initiate such calls absent an emergency purpose or exemption.
This seems a) obvious and b) not really big news. But the headline sells it well I guess.
Does that mean that the same restrictions hold for prerecorded messages and AI voice calls? That makes a lot of sense.
Does this have any impact on Google Duplex-like services? That was the thing that enabled Pixel users to ask Google Assistant to call a restaurant and make a reservation on their behalf, etc.
unsolicited. If the business has a contract with Google, Google can update the contact to say that they're allowed to.
That doesn't seem like a good idea. If Google/Microsoft really want to, they could get a big chunk of small businesses to allow them to do this. However there would be no way to build a competing service.
I feel like robocalls made on behalf of actual consumers in relation to actual b2c transactions should be allowed.
Immediate Reaction: FBI bans robbing banks while wearing woolen socks.
I wonder what qualifies as a robocall. Is it just something dialed automatically? Is it still legal if a human dialed the call, but an AI-generated voice speaks?
The law here bans both the use of autodialers and "artificial or prerecorded voices" in calls to cell phones (along with a variety of other types of phone numbers like emergency lines, other types of lines where you might pay for the incoming call, etc.).
Separately, it bans artificial/prerecorded voices in calls to residential lines.
Both provisions have carveouts for emergencies or when the party being called has given their prior consent.
This ruling was driven by fake Joe Biden robocalls, but there are/(were?) AI startups trying to create AI customer support bots or political reachouts with consent from the parties involved to clone those voices.
From the declaratory ruling, any AI-generated voice call requires prior recipient consent:
> Consistent with our statements in the AI NOI, we confirm that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that resemble human voices and/or generate call content using a prerecorded voice. Therefore, callers must obtain prior express consent from the called party before making a call that utilizes artificial or prerecorded voice simulated or generated through AI technology.
So presumably the google assistant "feature" that can book a table at a restaurant for you is now illegal? IIRC that would place a call to the restaurant.
This is a good outcome.
IANAL, but that would be the implication.
> AI startups trying to create AI customer support bots or political reachouts with consent from the parties involved to clone those voices.
This is where lawyers get to have fun. What is the line between a message in the public sphere copied and multiplied via broadcast, and a message consensually altered and multiplied via AI-then-broadcast?
The same law that bans artificial voices without prior recipient consent also bans recordings without prior recipient consent. So no difference whatsoever for phone calls.
Sounds like nothing of value is lost
This is a sad day for telephone scammer scammers.
Is it? I mean, the scammer scammers can still use AI to answer the phone. They just can’t initiate calls en masse using AI, which I don’t see them doing.
Some of them actively call in on known spammer numbers, like the numbers found on a fraudulent Norton invoice. Often the scammers wait for you to call.
Most probably will shrug their shoulders and say "well, anyways" while going about their regular scam calls.
you're talking about scammers not scammer scammers
I think they can safely assume they have a free pass here.
Looks like the FCC basically killed outbound AI calling companies like Air.ai, and does not seem to affect inbound companies like ours (https://echo.win)
Interestingly they explicitly mention AI generated voices, does that mean voices generated by traditional TTS engines are fine?
Those voices were already prohibited. This ruling specifically addresses agents "emulating human speech and interacting with consumers as though they were live human callers when generating voice and text messages".
Based on the (alarming) demo on Air.ai's homepage, that sounds like it would be prohibited unless the user consented to be contacted in that manner when providing their phone number.
So looks like the only allowed use cases will be for opt-in notifications and reminders.
How exactly do they propose to enforce this that isn't the same way they "enforce" already illegal robocalls?
I don't see why AI voices should be completely illegal in calls. Where I live businesses are required to disclose that a call is being recorded. I see no issue if they're also required to disclose that the voice I hear is AI driven.
That being said, robocalls are bs in general. What I'm saying is not an excuse for robocalls.
I don't know left from up in this situation, but I was under the impression outgoing calls are illegal, not inbound calls.
I don't see why it would matter for an end user answering or calling. I mean, the economics matter (a business can have way more AI voices than hired people to answer calls and send calls). But the experience of the human on the other end is probably ok if the human knows for sure it's an AI they are talking to.
I certainly close all those calls and not bother to interact with them regardless. But in terms of legality I would probably be fine with a restriction and not with an outright ban. Unsure.
Did they send a formal cease and desist to entire countries worth of scammers? Otherwise, this is yet another piece of feel-good legislation that will do nothing to stop my phone from going off.
Yes, I use RoboKiller. No, it doesn't stop everything. The text spam in particular has gotten crazy and it's not even close to election day.
I run a company that automates B2C sales lead follow on multiple channels and we use AI to leave polite messages for folks who consent based on their inquiry.
The problem we are solving is that about 1/3 of all web leads are fraudulent. Our clients are having trouble sorting through which leads are real people who want to do business and which ones are bots/BS. This ruling is disappointing.
There are better ways to solve this problem, as described for many years here and elsewhere there should be "postage" for messaging and calling. Sender pays, and they get their money back in full if the recipient responds. Costs spammers millions, costs normal people nothing or very close.
[dead]
https://www.wsj.com/tech/ai/fcc-bans-ai-artificial-intellige...
I just got an AI-generated voice call late last night about a missing elderly person in a nearby town.
The declaratory ruling reads such that a real-time translation service or artificial voice to accommodate a disabled live service agent would also be forbidden.
In both cases an artificial voice is being generated. This ruling seems to trample on some basic human rights.
Makes me think of the Google wavenet voices that pre-ceded much of this.
In the interim, this might be an understandable safeguard before elections while a clearer path forward is discovered.
I wonder if this will inspire the film industry with opposition to generative AI
Are there some realistic Text-To-Speech voice models out there that I can use locally and for free?
I know that ElevenLabs, Microsoft, of course OpenAI have some nice voices. But I would like to use them locally, or maybe in an app?
Would you support phones having an optional answering captcha system for untrusted numbers? Something like:
"answer the following question to complete your call: if Sally has two eggs and Michael has one, how many do they both have?"
My Dad's landline makes you press a digit before completing the call. So that exists already, and wish more would add this as a feature. I'm sure like anything, it could be defeated, if they had a system listening for the key to press. But it works for now.
Oh definitely, and and it will always be a game of cat and mouse.
That feature on your Dad's phone sounds like a decent step in the right direction.
I use google's call screening and it works wonders. https://youtu.be/V2IyttWHJfs?si=AW6fZQMl85w4srBM&t=48
To check the balance on a prepaid credit card I found on the ground (the modern equivalent of finding a $20 bill lol) I had to go through a prompt that said “press the number of the first digit of the following: eight, four, two”
So it works in some way for the CC companies at least.
Isn't that the sort of task that's easier for an AI than a human with other stuff on their minds?
I agree that it wouldn't catch all spam, but it might help reduce the amount of recorded robocalls waiting for someone naive enough to engage.
2 replies →
If you get the persistent scammer calls, you can transfer them to https://www.reddit.com/r/itslenny/
Yeah because other activities they have deemed illegal have totally stopped. I predict a season of AI generated robocalls for the elections. From all sides. This message brought to you by .......
Probably not very enforceable. There is already a case in Hong Kong where an employee transferred 25m to scammers because of a deepfake video call of scammers pretending to be his colleagues.
Perhaps illegal activity should be illegal, and AI-generated voices for legitimate uses must be allowed, otherwise this creates a prior restraint censorship situation.
How does this affect other countries dialing into the US?
In a prior actions the FCC cracked down on "gateway" phone companies that are known to connect lots of spam from abroad.
Did the crackdown measure as having worked? I don't know where to look up those stats.
2 replies →
I thought robocalls were already "illegal." Does this make them double bad? Is the FCC going to do twice as much nothing about the issue?
It's puzzling why wasn't this wasn't already illegal by virtue of robocalls themselves being illegal. Why are those allowed?
This is exactly what the ruling is doing. It is explicit confirmation that it was already illegal under existing law.
Seems like there's a potentially silly but also valid argument here that that's a literal violation of free speech?
As with everything, it’s all about enforcement.
People use phones to 'call' each other? When did they get this feature? Is it some variation of FaceTime?
I'm sorry, you must be trolling. It inconceivable, that in 2024, an audio to audio connection connection could be made between two "phone" users. What's next, phone numbers?
Does this mean all the AI voice assistant via phone startups are screwed ? Or it’s only for outgoing calls ?
I'm sure this will stop the 3-6 automated spam calls I get daily that originate outside the country.
phone calls as we know it are going to go the way of the dinosaurs, we need trusted communication systems
Agreed. Once mobile data coverage is universal (via starlink et al, maybe), it's inevitable that the idea of a phone number will become antiquated. Either whatsapp (or one of its competitors) gets a sufficient monopoly and enables easily portable identities (to allow switching sims), or some other similar platform will come along. It may take a decade or two, but it will happen.
But it's never going to be universal. I felt very scared some weeks ago during a huge march against rightwing extremists in Munich, Germany. There were ~150k people concentrated on a few streets/km.
Now, how is this relevant? Well, the entire cell network was offline, at least for some providers. At first it wasn't possible to send/receive data. Calls were connecting, but my friend sounded like an alien. Then for one hour, 0 communication was possible.
So even though the most efficient (I think?) protocol was used, it came to a halt
I used to think this about email also.
Pretty sure they can’t “rule” anything. There’s a few cases at the Supreme Court that should issue by April(?) regarding these agencies “legislating”.
I’m pretty confident this will not stand, for one, it violates the first amendment. You can’t tell anyone what messages, voices, thoughts, expressions, they cannot transmit. That’s been actually ruled in repeatedly
Deceptive trade practices are indeed illegal, and this tracks as the same kind of deceptive behavior.
Once again, red tape and bureaucracy are holding back productivity improvements.
(Only half joking here.)
Just like spam calls are illegal! Very confident people globally will follow our laws :)
Thank goodness. AI is already allowing enough manipulation of our elections as it is.
Wow, nice to see our government working for us and not just the corporatists.
Does this outlaw the Google thing that makes restaurant reservations for you?
Depends whether the restaurant asks "Which name do I put on the reservation" or "What's your name"
It takes three seconds of speech to generate a synthetic version. I think of my journey job searching and how much personal information I have to trust with basically random people. Voice, likeness, sample writing, resume. Everything is out there already but makes it a lot easier
Makes sense, impersonating people for gain and/or harm is illegal.
Why cant they make all robocalls illegal? Name me one good robocall
> Why cant they make all robocalls illegal?
Generally, speaking, the FCC can't pass laws, only interpret and apply them. In this instance they are not making a new law, they are declaring that the powers granted to them under the TCPA (a law passed by Congress in 1991) allows them to regulate/ban AI voiced calls.
> Name me one good robocall.
Government services. Voter info, school closures, water outages, etc.
The pharmacy calling to tell me that my prescription is ready. Those may be AI-generated too if they add the medication name in. Not sure if that's covered by this ruling though.
just rule robo calls illegal.
this is a baseless distinction.
if there's not a human on the other side, it's illegal. easy to prove, record a call, ask some dumb questions and all is simple.
this is a pointless line.
I'd much prefer they made it legal to brutally torture, rape, and murder these scumbags.
I suspect these things would completely end after 10 instances of the state getting out of the way and allowing nature to handle things the way it has successfully handled things for the entire history of humanity until very recently.
(Also, look at how old my account is and consider whether or not I care about your downvote. Reply with something that directly refutes the point I'm making so we can have the vibrant discussion this website used to be known for. Downvotes are simply pathetic attempts to silence correct views. Intellectuals have discussions, not censor their opponents. Only the most indefensible and mediocre positions depend on censorship and explains why the most unimpressive ideas depend on it.)
Of course regular robocalls are totally yeaaahy
That will do nothing to stop the Indian scammers
Oh yeah who's gonna enforce it? Hopefully they make scamming illegal too, it's utterly surprising they didn't outlaw it to prevent it from happening.
I don't think that will stop scammers!
why just not make all robocalls illegal instead :-) I dont care if its a AI voice or a recorded message.
Robocalls themselves should be illegal.
But how would the poor political campaigns reach all those uninformed voters? /s
The "poor political campaigns" already exempted themselves from needing to adhere to the "do not call" list. So were they to make robocall's illegal, the politicians would likely again exempt themselves from the "robocalls are illegal" law.
With the result that (assuming the existing robocallers all quit) the only robocalls one would get would be politician robocalls.
In any case, most all of the current robocalls are already "illegal" under one or more existing laws/regulations, yet they still occur because the ones making the robocalls face few (if any) penalties for violations.
Incoming pittance fine and a handie.
I'm braced..
> "State Attorneys General will now have new tools to crack down on these scams..." - FCC Chairwoman Jessica Rosenworcel
...How? How can this be enforced? What are the new tools? Based on the news release and documentation, fiat in this case means nothing but posturing, at most being hopeful some imaginary future tool will be able to bring execution to legislation.
Just ban robocalls
Then you turn plain old voice calls into an oasis for humans. Not a bad idea.
I hope someone is filing FOIA requests to get any communications between the White House and FCC related to this after robocalls were using Biden's voice. The timing here seems pretty dubious and as far as I'm aware the FCC is meant to be an independent body.
If the White House did pressure the FCC to implement a specific rule I'm pretty sure that would be a problem. The White House can obviously set general priorities, like protecting consumers from high fees, but isn't supposed to push specific rules, like requesting a new rule to ban hidden fees. PR in this case, if the White House specifically requested a ban on ML-generated robocalls that would be a problem as far as I understand it.
That is not going to stop republicans from spewing garbage.
maybe ban robocalls?
yay!!
[dead]
I'm so glad the FCC is protecting vital spam call center jobs /s
That's right. I want my robocalls to be human, like my granddad preferred. /s
[dead]
[flagged]
This website is reminding me more and more of libertarian facebook groups that I saw in the past. Goodbye Hacker News.
Please elaborate, for someone like myself who can't keep up with the latest belief systems of political parties?
Do libertarians have a strong view on this topic, and what is it?
Regarding the comments, I see very few inflammatory or divisive comments. The average comment here seems to be poking fun at the fact that robocalls are already illegal, and that banning the more specific "AI robocall" seems like security theater.
don’t let the door hit you on the way out
The most popular topics on HN are bureaucratic decrees. Sad.
Says the guy who submitted an HN post about an executive order, 29 days ago. (Which I see nothing wrong with, just pointing out the hypocrisy.)
Anyway, is it sad, really, for folks on hacker news to discuss regulations on information technology? Especially when the regulation pivots on, of all things, "AI"?
I think phone calls are dying, and the future of voice interface lies within apps, web, and new products like Vision Pro. At re-tell.ai, we aim to help developers create meaningful and humanlike conversations that will solve staff shortage problem, boost productivity, and unlock new opportunities. Check out our product hunt link: https://www.producthunt.com/posts/retell-ai
I suspect this will be challenged and the Supreme Court will overturn it on First Amendment grounds.
Why? Because creating hate and fear through variouis forms of media is a key part of politics. For example, local media (newspapers, radio and TV) are very big on ppushing crime hysteria narratives, despite crime being near all time lows.
There's too much vested interest in unlimited robocalls to let this ruling stand.
The one exception to all this is if you use an AI-generated voice to impersonate someone to say something they never said but this is already illegal on the grounds of defamation. The same applies to any deepfakes.
The real problem is that the phone network as it exists now needs to die. Add to that the decades-long effort to pack the court and overturn campaign finance laws (ie Citizens United v. FEC).
So I suspect this move will go nowhere. This will probably be even easier to challenge when SCOTUS overturns Chevron, as most expect them to do, essentially gutting executive agency power.