Comment by jillesvangurp
2 days ago
I think they are missing a few very significant revenue sources that likely going to grow beyond just simple enterprise user accounts for Chat GPT.
I'll call out two of them.
Image, video, and other content generation is going to become more important and companies will be spending on that. We've seen some impressive improvements there. IMHO near term a lot of that stuff might start showing up in advertising and news content, the whole media industry is going to be a massive consumer of this stuff. And there's going to be a lot of competition for the really high quality models that can be run at scale. Five years until 2030 is a lot of time for some pretty serious improvements to land.
Another area that the article skips over is agentic tools. Those are showing a lot of promise right now. Agentic coding tools are just the tip of the iceberg here. A lot of these tools are going to be using APIs. So API revenue is a source of revenue. There are applications across the entire IT industry. SAAS, legacy software, productivity tools, etc.
Yes 207B is a lot of money. And there's no guarantee that OpenAI comes out on top "winning" all these markets of course and we can argue about how big these markets will be. But OpenAI does have a good starting position and some street credibility here. It's a big bet on revenue and potential here. But so is betting against all that and dismissing things. And there's a lot of middle ground here.
> Image, video, and other content generation is going to become more important and companies will be spending on that. We've seen some impressive improvements there.
They are assuming that people _want_ automated content enough to pay for it.
Like at the moment its great because its essentially free, but paying >200billion?
Are we assuming that we can automate the equivalent of instagram, netflix/disney and whatsapp and make advertising revenue?
I think the unmentioned pivot is into robotics, which seems to have actual tangible value (ie automation of things that are hard to do now)
but still 200billion in r&d
Actually, I sense a mounting "AI fatigue". Angry comments under AI contents. Social media accounts that proclaim to be "AI Free". I was thinking today that the killer app for AI could be a filter that automatically exclude all AI content.
I agree. Back in the 90's there was plenty of skepticism, along with some mockery, surrounding the internet. But you never saw headlines like "Lawyer caught using the internet" or "Artist busted using the World Wide Web".
With AI, it's different.
Google are experimenting with different filters that intentionally blur the lines between real and fake video in their shorts. A creator I follow who makes human-in-shot videos suddenly looked fake, like some animation filter was applied. I commented on how bad it looked, and someone said YT are doing it to people's videos as some sort of "trial". The effects were to make the human look more AI generated. The only use for this I can surmise is to make AI content even harder to detect.
1 reply →
There's a huge vegan market segment but meat industry still plods along
3 replies →
It is very obviously there. People are not simply accepting the AI-bro justification that one day they won't notice the difference so it's all fine.
Ordinary people are understandably either exhausted by or angry at AI content, because it is relentless and misleading, and because they know instinctively that the people who want this to succeed also want to destroy employment and concentrate profits in the hands of a few Silicon Valley sociopaths.
> They are assuming that people _want_ automated content enough to pay for it.
People love money. That's not an assumption but established fact. Producing content manually is expensive.
You are way too emotional about this. Ads are a proven business model. We have AI models producing some pretty slick video and images now that are more or less publication ready quality. For next to nothing. The media industry not saving cost when they can is unlikely in my view.
Robotics is indeed another angle that OpenAI might do something with. I'd say that adds to my argument. That's a potentially really big market as well. But I don't consider OpenAI to be clearly leading that one. There are at least half a dozen companies active producing humanoid robots. But undeniably those might be a lot more useful with AI that understands and can interpret visual imagery.
> Like at the moment its great because its essentially free, but paying 200billion?
OpenAI needs to about 20-30x their current revenue over the next five years. A tall order. But these are large markets. It's not impossible. And when the financial times is ignoring major revenue sources it needs calling out.
Yeah, I have a hard time believing that there’s a massive demand for AI-generated videos and images. Like, why would the news industry want to generate images and videos with AI? It’s not news. The advertising industry maybe, but even then it’s probably not your top brands that go full in on it. If you see that all of Apple’s adverts are generated with AI, it’ll probably lower your brand perception.
We already put more importance on handmade goods vs. factory-made, even if the latter is cheaper and better quality. I have my doubts about humanity collectively embracing content generated from prompts by black boxes.
> Yeah, I have a hard time believing that there’s a massive demand for AI-generated videos and images. Like, why would the news industry want to generate images and videos with AI? It’s not news. The advertising industry maybe, but even then it’s probably not your top brands that go full in on it. If you see that all of Apple’s adverts are generated with AI, it’ll probably lower your brand perception.
(disclaimer: I don't work in ad and don't know more about it than the next person)
Whether the end product (the ad) is AI-generated or not is almost irrelevant. The whole production chain will likely be AIfied: to produce one ad you need to go through many concepts, gather reference images/videos, make prototypes, iterate on all that, and probably a ton of other things that I don't know about... The final ad is 1 image/video, but there's been dozens/hundreds of other images/videos produced in this process. Whether the final ad is AI-generated or not, AI will almost certainly (for better or worse...) have a major place in the production chain.
The cost of operating television studios and paying related staff, including on-air talent, is probably significant. I can easily see major news networks turning to AI-generated newsreaders. TikTok is already full of AI voiceovers; seems like a short leap, to me.
7 replies →
> We already put more importance on handmade goods vs. factory-made, even if the latter is cheaper and better quality
I guess if 'importance' means sentimentality, however, the "factory-made things" industry certainly makes the vast majority of the money to be made in manufacturing. Handmade is a niche. I think the argument is that whether people prefer it or not, AI-generated content will have a major place in the marketplace for content because of its natural advantages. In other words, a movie heavily generated with AI won't win any Oscars but there is a world where such things would make a lot of money and some of that money would flow to OpenAI.
I have a hard time understanding why I would want to waste my time watching something generated by AI when they clearly didn't value the time spent making it.
I have a limited lifespan, I'm not going to use that to consume slop.
1 reply →
Right, the main reason media has value is because of scarcity. Generated media is almost worthless because there's seemingly infinite piles of it.
If a painting costs $20 bucks, and you can generate 1 million paintings, that doesn't mean you just made $20 million dollars. It means paintings no longer cost $20 bucks.
> They are assuming that people _want_ automated content enough to pay for it.
I’m assuming gen ai will supplement so instead of ten artists producing content you’ll have three orchestrating and correcting but the output will be indistinguishable from what it was before. Just now the margin is much higher.
I never think of direct to consumer sales in this context.
Anecdotally I've seen lots of anti AI posts on instagram from gen-z accounts, and on YouTube ive seen "ai not used to make this music" in the description.
Also anecdotal, but I do think there's gen z anti-ai sentiment, these kids are joining a world where they will never own a house and have bleak prospects and now even art is something they can't do if AI takes off, so the market might not be there
Seems like older people love AI art, people over 50, which is a sizeable market don't get me wrong, but in 10-15 years idk
But whats the USP of that though?
> indistinguishable from what it was before. Just now the margin is much higher.
In both music and TV/film there has been an explosion in productivity. A kid in a bedroom with a mediocre microphone can produce something that sounds 95% as good as something that took a team of ten to do at a big studio. The same with VFX. If you compare "the golden compass" from 2006 to the series in ~2020 the quality of the TV show is much higher, at a higher resolution and has more VFX shots in it.
Both music and TV, the margins have dropped precipitously.
I just cant see how making it easier will bring up margins.
The whole VFX market is something like 10 billion USD a year.
Which is a huge amount, but if they only capture a piece of it, it might not amount to much compared to their spending.
Everything professionally produced you currently see uses a lot of CGI, and you can’t even tell.
Essentially what they are doing is cheaper, more accessible CGI. And in the same way at it is now, you are not going to be able to tell it was used in expensive productions and will be able to see it in the cheap productions.
As an ex VFX infra guy, I can totally tell.
But to your point, if we take all of TV production world wide I doubt thats going to add up to 100 billion spent on set extension/characters/de-aging/fuckit fix it in post. And thats keeping spend at the same rate. (not to mention the recent peak spend on TV)
For openAI to make money it has to be an order of magnitude cheaper, quicker and quality, and be the lead so that people spend AI bucks with them. Rather than having a VFX company tune an opensource/weights model.
I would assume that generative AI will be able to make endless hours of personalized reality TV content that will hit a surprisingly large demographic. And then it can use recommendation algorithms to distribute that reality TV AI slop to all the people that might or might not want it in order to turn it profitable. Evidence: all the god-awful foreign reality TV on Netflix currently.
It's a bit harder to replace Disney at this point, but give it 20 years. I'm bullish on AI slop on par with The Apple Dumpling Gang or Herbie Goes to Monte Carlo within a decade or so. Evidence: The rapid evolution of AI generated video in the past few years extrapolated into the future.
I keep thinking about how, 20 years, 3D printers became accessible for tech/price, and there was a lot of talk about how we'd all just print what we need (pretty much like in Stephenson's book Diamond Age), and you'd get rich selling digital patterns for material goods.
Instead, we got the elevation of the handmade, the verifiably human created, typified by the rise of Etsy. The last 20 years have been a boom time for artists and craftspeople.
I keep seeing AI slop and thinking that all this will do is make verifiably human created content more valuable by comparison, while generative AI content will seem lowbrow and not worth the cost to make it.
2 replies →
> IMHO near term a lot of that stuff might start showing up in advertising and news content, the whole media industry is going to be a massive consumer of this stuff.
One of the candidate for mayor of New York ran a whole AI video ad campaign.
Half of the real estate listings have AI remodeled pictures.
Probably a quarter of the printed ads I see each day are AI generated.
It will only increase, but it's already here.
Real estate agents only use AI because it's cheaper than paying $70 CAD per image for human-made images. There's a cap on how valuable the AI generated images are.
The last time I sold a house, a professional photographer took some beautiful photos, gave the realtor like 60-80 and he picked 40 for the listing. He absorbed the cost of that with his commission. No way in hell did he pay $2800-5600 (I'm also in Canada) for those photos.
1 reply →
The main thread through the case laid out has a fundamental flaw "assumes a spherical cow in a vaccuum" . There are cheaper/faster more open competitors out there doing as much as OAI and getting integrated in real product pipelines across the planet.
Its not a Blueray vs HD-DVD situation (winner takes all and competition dies on the vine) or early Google vs Yahoo Search i would posit its more like bitkeeper vs git - and will probably eveolve into a git/hg/bazaar type situation. Costs will only keep dropping and the easiest/cost effective options will get faster uptake in the backend dev world where it matters.If im integrating a LLM and the whole field can kinda do the equivalent capex/opex will push the decision.
I think actually selling media is going to be tough. 1: you have to beat the current price point so people will use it, 2: it can’t be too cheap or you can’t make money. That mad market price point makes it hard to get right. The more content the fewer eyeballs, meaning the less value, meaning people willing to pay less. If it makes making movies a lot cheaper, then there’s a lot less money to be made.
The thing about tech is it has to scale to meet tech valuations, and there’s a reason instagram basically gives content creation tools away for free.
These are all commodity use-cases though. Google, Meta, and Anthropic already all have competing products of equivalent quality and customer pricing is being driven down aggressively.
But there's fierce competition since no moat. How will generated images and videos avoid being a stock photo situation?
The moat is the infrastructure needed to run this at "Google" scale for the entire planet. So far only OpenAI and Google have managed to pull that together. It's just really hard to raise the hundreds of billions you'd need for that.
This sounds very similar to the Tesla stock value hype train. It’s all a gamble on the promise of tech that isn’t quite there yet.
What I don't understand is how Elon will ever compete with Chinese humanoid robot pricing and output. He's known for not caring about privacy, so there's no real difference between buying a Tesla one or a Chinese one. Both will look and listen into your home and transfer as much data as possible to their true owners.
He congratulates Jeff Bezos for New Glenn, and rightly so, but he's absolutely silent about the Chinese robots. One has to admit that they're already ahead of the US. This comes from a German who well recognizes how far behind Germany is in comparison to the US, in regards to any digital technology.
Oh that's easy: as soon as China gets good enough to compete at any market seen as vital, "the west" either makes up some nebulous national security threat to cut them off or imposes high enough import taxes to make the home-made versions cheaper in comparison. The same way "the west" has done it with cars, phones, and social media in recent years.
There is no free market competition between "the west" and China, they're two completely separate markets for a reason. You're only allowed to cross that isle in a limited capacity, as soon as you get big enough to be seen as a legit threat, you're getting cut off. Works both ways, Tesla can sell in China, but they can't beat BYD. Apple can sell iPhones, but only because their market share is not big enough to matter.
3 replies →
Probably China is building useful robots, not gimmicky "humanoid" ones. For any given task or set of tasks I'm very sure there are vastly better robot designs than mimicking a human.
> how Elon will ever compete with Chinese humanoid robot pricing and output?
... with said humanoid robots?
While it's true China's manufacturing sector can outproduce the west's right now, ubiquitous autonomous humanoid robots can level the playing field to a meaningful extent.
1 reply →
Yeah, considering Teslas upload imagery from their cars to datasets their employees have been known to access (and abuse that access as well) there's practically 0% chance I'd ever let a Tesla camera inside my home. Its bad enough they're recording me when my neighbors drive by and when I walk around, I wouldn't want them in my home on a device that's designed to be able to go through my stuff.