Comment by scoofy
9 days ago
So, I really felt like more people should be reading Nassim Taleb's Incerto series of books. A lot of the issues that fall out of AI he dealt with in his books like ten years ago.
He gives one the best pieces of advice I've ever heard: if you are going to do something for a living, make sure it is NOT scalable.
If you do something that isn't really scalable, like being a welder or a tailor, then you only have to compete against the tailors in your neighborhood, and you can easily find a neighborhood that doesn't have a tailor. If you're building a scalable product, you'll always be competing against the best, most well funded, smartest people in the room.
Everyone here has grown up in the birth of the internet -- a once in ever event -- where building something scalable was just there for the taking. That's never going to exist again basically.
That seems like reasonable advice until you realize you have no idea what will and will not scale in a few years, and there's only so many tailors/plumbers/welders/etc.
What it really comes down to is these options as more work gets automated:
1) New jobs doing different things that weren't done before
2) Same jobs but shorter hours so "full time" with a salary to match starts to look more like 4 days, or 5 hours/day, or something.
3) Lots of unemployment
These can happen in a lot of different combinations, they can come wrapped up in different ways, and unequally for different segments of the workforce, but there's limited elasticity in most areas where additional people piling into the field would create more demand rather than glut the supply to the supplier's detriment.
I keep hearing about the potential of "new jobs" coming from ai but can anyone actually describe one? My gut says they will be something similar to converting middle class knowledge workers to do DoorDash drivers or trained artists becoming dog groomers. What a cool future, at least my parent's can watch racist ai slop videos on their iPads.
[dead]
It's simply a case of looking back and deciding this technical revolution is identical to the ones preceding it. Thus jobs destroyed will be replaced with new jobs, it always happens that way.
Of course past performance is no guarantee of future success...
8 replies →
The point of doing something non-scalable is that you can enter and exit the market fairly easily. You don't need to be a tailor your whole life. You can make a living as a barber, electrician, teacher, or nurse.
I'm not saying it's easy! It's hard as hell. It sucks when your job gets automated. I'm just saying that aiming for something non-scalable means you're not always tilting at windmills, and the game can't be rigged against you.
Switching jobs from electrician to teacher to nurse will take around 3 to 5 years of education or apprenticeship in most countries. It also requires new licenses or certificates if you ever move country.
5 replies →
All of those professions you've listed require about half a decade of dedicated training to be legally allowed to practice. For example an electrician takes like 7 years to become qualified, that's a full time apprenticeship, and it pays badly in the meantime.
A fact endlessly annoying to electrical engineers who legally can design their houses power system but not work on it.
(I mean I think a barber is quicker, but one of that list is also not like the others)
4 replies →
I used to look at my local dry cleaner and think that it had to be the most stable business going - it couldn’t be replaced by a computer. Then Covid hit, work from home took off, and the dry cleaner went out of business. Computers came for the dry cleaner, not by cleaning clothes but by eliminating the need for dress codes.
Being a tailor is scalable, that's way there are way, way more cheap machine produced clothes today than in the past. Surely he did not miss that the textile industry was at the core of industrial revolution. So being a tailor is more like a post-scaling job - the automation has already happened and now there are only remnants left.
But how can you be sure a job is peak-automated? A few years ago, I would have said musicians are post-scaling - way fewer musicians jobs now that you can play recorded music. But it looks like generative AI will hit musicians again. Can some of welding be automated? Probably.
>how can you be sure a job is peak-automated?
Probably the best way is to spend a few years working for a company where you can get a better picture whether it seems that way or not.
I mean a tailor who adjusts clothes and occasionally makes something bespoke.
Tailors typically operate a launderette and act as middlemen to a local dry cleaner.
I’m not talking about a fancy man making clothes for rich people, I’m talking about the talented old lady in you neighborhood who adjusts your clothing for $50 and runs a wash and fold.
>Can some of welding be automated?
Huge amounts have been doing it for decades.
Manual work pays better than ever though.
And plenty of alterations going on all the time after all the automation dust had settled manufacturing most fashions, a lot less manual work is of course being done but it's still everywhere. You do have to be good or you're not going to do half as well as you could though.
The thing is, automation should be expected to slow or stall sooner or later, automation's not suitable for every little bit of welding or sewing that needs to keep going on. Only the most suitable, of course ;)
These are just random examples, if you want to make absolutely sure you won't be automated away by the internet, build a valuable skill that doesn't depend on the internet at all, nor look anywhere near the places where automation is emerging that it wasn't doing before.
If you eventually figure out how to automate that skill it would be something.
Just like the internet though, there can be extra credit for being first :)
One of the most valuable things to be able to build single-handedly is something that can not be mass-produced by any stretch of the imagination.
You might stick with that alone, or pivot to something with more of a financial upside, but you would always have something to fall back on if needed. Plus give you less worry about taking financial risks than you would have been, considering the same resources and/or capital to work with.
And on a regular basis revisit how far you can stretch your imagination to see if your baseline fallback still doesn't look like it will ever be automated in a way that would effect you.
Yes, but this is simply the remnants of the old tailor occupation, post automation. The talented old lady would have had a lot more business in clothes making in the past, no need for a wash & fold.
My mom altered clothing when I was younger - and she darned socks too. My M-I-L still sews the occasional seam for pants that are too long and were cut for my wife or daughters.
But me? I buy a pair of $30 jeans at Costco. If they don't fit great, I buy a different pair of $30 jeans. I don't spend $50 to have them altered, or take it to a laundrette. If it can't be washed in our home washer/dryer, I don't buy it. And these days, when a sock gets a hole? I throw it away.
3 replies →
How many people visit either kind of business these days? I'm almost 35 and have never once gotten any garment tailored. I think one reason is that clothes are so cheap you can just keep looking and you'll find something that fits, which seems to be what most people do nowadays.
1 reply →
Welding is being heavily targeted for automation, apart from pressure vessels etc, most welding can be automated now a days , very soon ( months? ) every welding can be automated.
You're thinking of factory welding, manufacturing or maybe repetitive pipeline welding and thinks like that.
It'll be a while until a robot disassembles a trailer enough to remove the bent axle, cleans off all the paint and rust, bends the trailer back into shape where needed, cuts a custom support to makeup for some lost strength, welds it in, primers it, paints it, and assembles the trailer.
Same for construction too.
3 replies →
> if you are going to do something for a living, make sure it is NOT scalable
You need to consider both horizontal and vertical scaling. Being a bespoke tailor may not scale vertically, but it can scale horizontally. If you have too many people pick up tailoring, you might run out of neighborhoods without competitors.
> if you are going to do something for a living, make sure it is NOT scalable
I guess it also means that if you build something for a niche audience then big companies will never be interested in it.
That's not really a bad thing, IMHO. Many people successfully make a living creating niche products.
I wouldn't say it's not really a bad thing, I think it's a very good thing. There are many people now making incredibly niche products that have very good lives - making more than enough money for themselves doing interesting work engaging with customers that are passionate about their field.
Sounds like a great life to me.
I like doing things for big companies that they actually could afford easily, but their costs are so high it's a better deal to have me do it.
One of my favorite niches :)
Some clients are good enough at math to figure it out, and you don't actually need that many of them if they are big enough.
how do you approach big companies to offer them these custom niche services?
1 reply →
What are the things you do?
1 reply →
I'm sure someone else on Etsy will undercut your price if you sell more than a few.
Probably until a copyright vulture comes along and nukes it.
1 reply →
Welding is coming along. Ever heard of Path Robotics? That's high-mix, high-complexity welds. (There are a lot of other, less-sophisticated welding robots out there.) The biggest moat for a welder right now is special certifications to be able to do welds on submarines, etc.
> He gives one the best pieces of advice I've ever heard: if you are going to do something for a living, make sure it is NOT scalable.
You mean like opening a restaurant?
Too bad delivery services like Uber Eats totally own the market now.
Starting a hotel? Booking.com and Airbnb are there to take your profit margins away!
I think that delivery services give you bigger market, but not intrinsic scale. You're still limited by kitchen size, staff numbers, and raw hours you can put into the food.
You can scale the system (say Subway, or even smaller chains like Burger Fuel), but also reasonably choose _not_ to scale and still do incredibly well (like Michelin star restaurants, or the myriad of hyper-famous-locally Japanese eateries, or Fergburger in Queenstown).
Someone scaling their own restaurant on the other side of the world won't necessarily out compete out (and in the overwhelming majority of cases with have no impact at all). Despite fast food joints all over the place, I still see small cafes, individual eateries, etc performing well (I mean, as well as hospitality can be).
Maybe it's worth expanding the definition of > make sure it is NOT scalable < to include 'make sure it's not automatable'?
The problem is that any market can be taken over by someone with deep pockets (or investors making a pile of money). They just make sure they are the "go-to place" for consumers to access the market. By marketing the hell out of it, by making apps, abusing power in other markets (see platforms), etc.
Yes and any other business has to play the Google Maps and SEO game.
That's also the advice of the founder of this website, but not his investment firm since it's looking for moonshots. https://paulgraham.com/ds.html
I'm excited for a future where the technologist is like the tailor in their community. Scaling software has created a host of 'product traps' and there is no need for that for the vast number of activities people do aided by technology.
New laser welders from xTool lower the barrier to entry for welding to nearly the floor. They basically automatically push the welder tip along at the perfect rate for excellent welds, and get super deep penetration / full fusion welds every time with very little skill.
It will be difficult to know what career is ‘safe’…especially if everyone floods into it because it’s one of the last decent-paying jobs left (it wont pay well for long if everyone else loses their jobs).
It's so interesting the amount of people with these big AI fears who think that AI is going to replace most knowledge work within a short period of time, singularity, etc., but that same AI that takes over everything .. isn't going to be smart enough to operate robotics to do plumbing or welding? Those things will be outside the limits of its intelligence?
how good was Taleb at following his own advice? Had he tested it? As I recall he is pretty big on "skin in the game' as his differentiator.
He kind of did. He used leverage to place bets on the market.
Leverage scales in a way, because larger pools of money (corporations, groups of investors) mean larger bets.
But leverage also doesn't scale, because larger pools of money tend to make smaller bets over a larger surface of the market, since being wrong in a big way with leverage can wipe you out.
stupid advice. only way to solve the problem is collective. even if you become a tailor or a welder, what about other people? it's not a singleplayer game at all.
>> you'll always be competing against the best, most well funded, smartest people in the room
What if your goal is for them to buy you out?
There are much less expensive ways of playing the lottery than investing a year of you life into an unprofitable product that must be scaled to be successful, and only then hoping gets bought by a direct competitor.
Don't compete, complement and start building the relationship from the beginning.
Plus, who wants to spend any time competing with the stongest contender anyway, especially to get started :)
Is providing scalable products at a cheaper price scalable? If so, can it hurry up and scale? This is a bit of a paradox.
Who cares? He's arguing about providing something that is not scalable: so whatever happens to things that are scalable ain't the topic.
automation makes every job scalable, no?
Even if it does, there will probably be a prolonged economic period where robots are doing dangerous/messy stuff like welding, plumbing but there is a human master guiding them from a few yards away, via prompts, controllers, etc. More of a semi-autonomous power tool than an fully autonomous master that is delivered by drone on-demand. Scalability is still a ways off.
If we get AGI and fully autonomous robot assistants, we'll live in a post scarcity world like Star Trek, or somebody in control the robots will use them to enslave all of humanity... so... high variance outcome could go either way.
given the history of humanity, the post-scarcity world of Star Trek seems highly unlikely
That is the ultimate goal. We are very far from it for many jobs.
Do you expect we’ll have AGA, artificial general automation?
I don't think it needs to be general to be scalable to the point that it causes previously safe industries to "race to the bottom"
Uber didn't need to launch in every neighborhood to start destroying the taxi business
Facebook didn't need to have every eyeball on the internet in order to massively disrupt trad advertising
At some point there's just a tipping point
I do. But it will take some time to get there.
I don’t see anything stopping that from happening in the very near future.
1 reply →
could you automate a BJJ coach?
Who's going to pay for a BJJ coach when nearly everyone is out of job?
1 reply →
IMO literally automating the job isn't the only possible scenario -- already in the weightlifting space coaches are supplying skills and system prompts to llms that do their training for them, massively raising the number of students they can train at at time -- at some point that turns training into a zero sum game
I mean, if AI doesn't stall out, maybe there will be BJJ bots in 10 years.
can you automate a tailor?
Isn't tailors by and large already outcompeted by cheap new clothes?
13 replies →
Absolutely. Design parametric families of patterns, 3d-scan the person, let customer adjust with live preview, laser cut, then fully automated or low-skill assembly. Probably not currently economical like many things involving physical world manipulation, but without obvious roadblocks.
yes but is there enough demand for tailoring that it would make it economically feasibly to create the investment to automate it? probably not.
and in a world where many/most(?) people have lost their jobs to AI, only the wealthy few will be able to afford tailoring anyway
This seems overly pessimistic
As scalability becomes more accessible like with coding agents
The less it becomes about money but distribution
While money can buy you the latter it may not compound to something that is sustainable
> if you are going to do something for a living, make sure it is NOT scalable.
Great advice but difficult to action though.
I mean 10 years back I'd have thought programming is that thing which is not scalable. I had every reason to believe that. It required skill, experience, ability to stay current, grit for debugging hard stuff. Much of it can be automated now.
What can I pick now for a living that is not scalable today that some future technology would automate it just as easily.
You never wrote any scripts to write code?