Maybe this is alarmist, but I don't see how LLMs don't collapse our entire economic system over the next decade or so. This is coming for all of us, not just the NLP experts in big company research groups. Being able to cheaply/instantly perform virtually any task is great until you realize there is now nobody left to buy your product or service because the entire middle class has been put out of work by LLMs. And the service industries that depend on those middle class knowledge workers will be out of work because nobody can afford to purchase their services. I don't see how this doesn't end with guillotines coming out for the owner class and/or terrorism against the companies powering this revolution. I hope I'm wrong.
The owner class gets enlightened and makes sure that the govt taxes them and implements a solid Universal Basic Income
This is part of what the original UBI concept was about.
If this doesn't happen, yes, there will likely be violence until it is fixed.
The other view is that many technologies that were supposed to reduce work actually net added work, because now more sophisticated tasks could be done by the humans, so the net was similar to the highway paradox where more and wider highways breed more traffic by induced demand.
Where would this demand come from? IDK, but at least initially, these LLMs make such massive errors that keeping a lid on the now-hyper-industrial-scale bullshit[0] spewed by these machines will make many more full time jobs.
Seriously, just today I was amazed at how the GPT model tried to not only BS me with completely fabricated author names for an article that I had it summarize, but it repeatedly did so even after being successively prompted more and more specifically to where it could find the actual author (hint: right after the byline starting with the word "Author". It just keep apologizing and then doubling down on more fantastic lies, as if it were very motivated to hide the truth (I know it's not, that's just how fantablous it was).
[0] Bullshit being defined as speech or writing telling a good tale but with zero regard to the truth or falsehood of any part of it — with no malice but nonetheless a salad of truth and lies.
The answer will either be some socialist rationing policy, or genocide.
Current "work for a living" systems only sustained the population because a human could be the most cost-effective way to get something done. Unless there are still tasks where human labor is the best option (research jobs maybe), this entire economic system will collapse.
> The owner class gets enlightened and makes sure that the govt taxes them and implements a solid Universal Basic Income
But where will they make their billions from if everyone will be living on a basic income? Less money for them will mean less tax money, and less UBI. It will spiral out of control into complete societal collapse if AI doesn't hit a plateau soon.
That is why I said "solid UBI", as in more than merely survival wages, i.e., enough to not merely buy food & shelter, but also to live.
That said, this does need some thinking through multiple stages. On one hand, societies did still work when there was effectively unlimited slave labor, but that may be no more than a rough proxy.
Go to the endpoint assumption that AI and robots can produce everything needed for the population to live, and they are owned by 1% of the population. They made so much money so fast that they bought all the means of production. As of 01-January-2025 everyone is fired. Now what? Your'e right, no one can buy anything. The remaining populace cannot do anything because the new oligarchs have enough money and power to buy &/or threaten any politician.
The population overall is not going to simply lie down and die. About four days after running our of all the food in their pantries, they'll be revolting in the streets. One plausible result is a lot of carnage and the oligarchs are all killed and the AI is destroyed and outlawed. Or, they actually have sufficient command of the military and no defections and the military isn't smart enough to figure out that they're next on the starvation list, so the populace is wiped out as they revolt, and the world is left with the 1% of owners, and 1% of military. Or, there's some kind of balance reached, and the non-AI-owning class of fired people reconstitutes something similar to last year's economy, while the 1% go off to Mars or withdraw into their metaverse-ish thing.
That's just a few random thoughts on rolling the dice among the big forces, but it never plays out like that, so all of these are 99%+ likely to be wrong.
Seems like the only thing we know is that this potentially massively magnifies instability.
There are entire sectors of the economy that LLMs can't touch - hospitality, manufacturing, caregivers, religious sectors, live-action entertainment, etc. Sure some of these will be replaced by robots but there will always be new jobs too.
No, there are not. Everything in the economy is connected and you can't have a vibrant industry without customers. The customers of hospitality/entertainment/healthcare/etc businesses are largely the middle class who will be put out of work by LLMs. So the person who today makes $200/night in tips waiting tables at a nice restaurant.... who will be buying those meals?
I think the material point is that there will be far fewer of those people if everything goes well. You should need fewer people to do any arbitrary task when you are leveraging LLMs.
My own opinion is that people are going to have to become creators. And quickly. You can still create digital products, but you'll need to be a lot more quiet about what you're doing. And you'll need to have a facility for abstract thought to come up with ideas that no one else has yet.
With a few exceptions, using an LLM to perform a useful service is something that almost anyone will be able to do. Therefore these jobs will be not pay well.
That seems kind of like saying "using Excel to add numbers is something that almost anyone will be able to do" -- true, but the difficult part is (obviously a vast simplification) determining which numbers to add, under what conditions, and to decide what to do based on the result.
There is huge variation in how well people can prompt LLMs, prompt-engineering has many tricks that aren't obvious, that's why there's prompt engineers,
the only reason they studied, went to university etc was to avoid doing manual labour. this has been happening for decades, a century. they ll be depressed
Just give them the same lecture they like to trot out about supply and demand and how automation simply creates new opportunities. And then have an AI compose a dirge to play on the world's smallest violin for them.
Yeah also considering the article on Ars today about people testing whether ChatGPT poses no harm. I think singularity is coming but much differently than anticipated. So yes, the economy will have to change massively and less work hours are needed. (And the concepts are there: Negative Income tax, Universal basic income...) The change will be more graceful the less everybody is in denial.
As mentioned elsewhere this is not the first technological disruption in the economy. The change from heavy industry to a service industry didn't go well, hopefully it's possible to take learnings from this and do it right this time.
I'm not completely sure but considering how the coal industry went slowly out-of-business with subsidiaries going on for decades - at least that's how it was in Germany - this doesn't seem the way to go. At least I'm more than surprised that companies and the state were looking into early pensions instead of options to start from 0 at a new job.
For other legacy industries like the German ICE car industry it was at times a close call, so that's when the 35 hour week became widely adopted during the 90s. Even today it is still an option for new people joining (working on non-legacy products of course).
