2) I do see this, given the money poured into this cycle, as a potential possibility. It may not just be LLM's. To another comment you are betting against the whole capitalist systems, human ingenuity and billions/trillions? of dollars targeted at making SWE's redundant.
3) I think it can disrupt only knowledge jobs, and be some large time before it disrupts physical jobs. For SWE's this is the worst outcome - it means you are on your own w.r.t adjusting for the changes coming. Its only "world changing" as you put it to economic systems if it disrupts everyone at once. I don't think it will happen that way.
More to the point software engineers will automate themselves out before other jobs for only one reason - they understand AI better than other jobs (even if objectively it is harder to automate) and they tend not to protect the knowledge required to do so. They have the domain knowledge to know what to automate/make redundant.
The people that have the power to resist/slow down disruption (i.e. hide knowledge) will gain more pricing power, and therefore be able to earn more capital taking advantage of the efficiency gains made by jobs being redundant from AI. The last to be disrupted has the most opportunity to gain ownership of assets and capital from their economic profits preserved. The inefficient will win out of this - capital rewards scarcity/people that can remain in demand despite being inefficient relatively. Competition is for losers - its IMV the biggest flaw of the system. As a result people will see what has happened to SWE's and make sure their industry "has time" to adapt particularly since many knowledge professions are really "industry unions/licensed clubs" who have the advantage of keeping their domain knowledge harder to access.
To explain it further even if software is more complicated; there is just so much more capital it seems trying to disrupt it than other industries. Given IMV software demand is relatively inelastic to price due to scaling profits, making it cheaper to produce won't really benefit society all that much w.r.t more output (i.e. what was good economically to build would of been built anyway in an inelastic demand/scaling commodity). Generally more supply/less cost of a good has more absolute societal benefits when there is unmet and/or elastic demand. Instead costs of SWE's will go down and the benefit will be distributed to the jobs/people remaining (managers, CEO's, etc) - the people that dev's think "are inefficient" in my experience. When it is about inelastic demand its more re-distributive; the customer benefits and the supplier (in this case SWE's) lose.
I don't like saying this; but we gave AI all the advantage. No licensing requirements, open source software for training, etc.
What happens when a huge chunk of knowledge workers lose their job? Who is going to buy houses, roofs, cars, cabinets, furniture, amazon packages, etc. from all the the blue-collar workers?
What happens when all those former knowledge workers start flooding the job markets for cashiers and factory workers, or applying en masse to the limited spots in nursing schools or trade programs?
If GPTs take away knowledge work at any rate above "glacially slow" we will quickly see a collapse that affects every corner of the global economy.
At that point we just have to hope for a real revolution in terms of what it means to own the means of production.
- The people that are left and the people that it doesn't happen straight away on: i.e. the people who still have something "scarce". That's what capitialism and/or any system that rations resources based on price/supply/demand does. This includes people in things slower to automate (i.e. think trades, licensed work, etc) and other economic resources (landowners, capital owners, general ownership of assets). Inequality will widen and those people winning will buy the houses, cars, etc. The businesses pitching to the poor/middle classes might disappear. Even if you are disrupted eventually being slower to disrupt gives you relatively more command of income/rent/etc than the ones disrupted before you giving you a chance to transition that income into capital/land/real assets which will remain scarce. Time is money. AI is a real asset holder's dream.
- Unskilled work will become even more diminished: A lot of people in power are counting on this to solve things like aging population care, etc. Move from coding software to doing the hard work in a nursing home for example is a deflationary force and makes the older generations (who typically have more wealth) even more wealthier as the effect would be deflationary overall and amplify their wealth. The industries that will benefit (at the expense of ones that don't) will be the ones that can appeal to the winners - resources will be redirected at them.
- Uneven disruption rates: I disagree that the AU disruption force will be even - I think the academic types will be disrupted much more than the average person. My personal opinion is that anything in the digital world can be disrupted much quicker than the physical realm for a number of reasons (cost of change/failure, energy, rules of physics limitations, etc). This means that as a society there will be no revolution (i.e. it was your fault for doing that; why should the rest of society bear the cost? be adaptable...). This has massive implications for what society values long term and the type of people valued in the new world as well socially, in personal relationships, etc.
i.e. Software dev's/ML researchers/any other white collar job/etc in the long run have shot themselves in the foot IMO. The best they can hope for is that LLM's do have a limit to progres, that there is an element of risk to the job that still requires some employment, and time is given to adjust. I hope I'm wrong since I would be affected too. No one will feel sorry for them - after all other professions know better than to do this to themselves on average and they have also caused a lot of disruption themselves (taste of their own medicine as they say).
2) I do see this, given the money poured into this cycle, as a potential possibility. It may not just be LLM's. To another comment you are betting against the whole capitalist systems, human ingenuity and billions/trillions? of dollars targeted at making SWE's redundant.
3) I think it can disrupt only knowledge jobs, and be some large time before it disrupts physical jobs. For SWE's this is the worst outcome - it means you are on your own w.r.t adjusting for the changes coming. Its only "world changing" as you put it to economic systems if it disrupts everyone at once. I don't think it will happen that way.
More to the point software engineers will automate themselves out before other jobs for only one reason - they understand AI better than other jobs (even if objectively it is harder to automate) and they tend not to protect the knowledge required to do so. They have the domain knowledge to know what to automate/make redundant.
The people that have the power to resist/slow down disruption (i.e. hide knowledge) will gain more pricing power, and therefore be able to earn more capital taking advantage of the efficiency gains made by jobs being redundant from AI. The last to be disrupted has the most opportunity to gain ownership of assets and capital from their economic profits preserved. The inefficient will win out of this - capital rewards scarcity/people that can remain in demand despite being inefficient relatively. Competition is for losers - its IMV the biggest flaw of the system. As a result people will see what has happened to SWE's and make sure their industry "has time" to adapt particularly since many knowledge professions are really "industry unions/licensed clubs" who have the advantage of keeping their domain knowledge harder to access.
To explain it further even if software is more complicated; there is just so much more capital it seems trying to disrupt it than other industries. Given IMV software demand is relatively inelastic to price due to scaling profits, making it cheaper to produce won't really benefit society all that much w.r.t more output (i.e. what was good economically to build would of been built anyway in an inelastic demand/scaling commodity). Generally more supply/less cost of a good has more absolute societal benefits when there is unmet and/or elastic demand. Instead costs of SWE's will go down and the benefit will be distributed to the jobs/people remaining (managers, CEO's, etc) - the people that dev's think "are inefficient" in my experience. When it is about inelastic demand its more re-distributive; the customer benefits and the supplier (in this case SWE's) lose.
I don't like saying this; but we gave AI all the advantage. No licensing requirements, open source software for training, etc.