The legal work stuff was just to illustrate my point. I can only estimate as an outsider on how the legal profession will change in the long run as LLMs take over.
As far the research on LLM use in productivity goes, the mostly consistent finding is that low skill workers benefit the most and they can use it to catch up with high skill workers. LLM use didn't seem to benefit high skill workers as much suggesting that high skilled workers are the ones who're the mostly likely to lose in the adoption of AI.
I wrote a short note about the potential impact on the labour market. We mostly use our brain powers for job stuff instead of intellectual pursuits. The job market impact is more relevant for most people.
Interesting point. I believe you are suggesting that there will be a shift in IQ distribution, as LLMs enable individuals to qualify for jobs that previously required higher IQs (using IQ as a rough proxy for intelligence or ability to do stuff). For example, someone that couldn't be a paralegal before now can, but it seems this breaks down eventually. Will current paralegals now become lawyers?
This situation appears to create a bottleneck in various professions. If, in my example, if the current paralegals can't jump to lawyers, we now have more people that can be paralegals with LLM suppressing the wages of paralegals.
I don't think it will work this way. In my example, the current paralegals will be able to get more work done, and the "lower IQ" people still won't be able to compete. In the end we'll need fewer paralegals. In this case the paralegals will not necessarily use more neurons as they manage the LLM more than writing and thinking about briefs.
We'll see, but the time-use data suggests that if folks have more free time, they aren't going to use it for intellectual pursuits. If presented with the choice of a cookie or broccoli, most will choose the cookie. We have already seen that, given the choice of a book or a screen, most choose the screen.
The price elasticity of demand governs this. If dropping the price of legal advice to a tenth of current prices increases demand for legal services a hundredfold, then we could expect employment to go up. Some of the new employees will necessarily have "lower IQ" than the existing ones.
That level of price elasticity seems plausible for legal services - most people outside of large wealthy corporations use lawyers only rarely. It seems like there is a lot of latent demand priced out of the market.
Of course the lawyers' unions (bar associations?) could frustrate this by colluding to keep prices high.
I'm not sure that AI drops the price of legal services. I still think there is a bottleneck issue. Yes, paralegal work will get cheaper, either by more paralegals getting paid less while using AI or super paralegals that can do more with AI, so we need fewer of them. In the end, all legal work has to be approved by lawyers. Will lawyers use AI to proof the AI work done by the paralegals? That seems risky, at least for now. Even if paralegals can produce more of their work for less, lawyers can't (so far).
I think this goes for various professional activities; as long as someone has to review the work done by AI then prices don't change much.
Predicting the future is a risky endeavor. Either way, AI will be disruptive, but I don't think it is clear yet exactly how.
The legal work stuff was just to illustrate my point. I can only estimate as an outsider on how the legal profession will change in the long run as LLMs take over.
As far the research on LLM use in productivity goes, the mostly consistent finding is that low skill workers benefit the most and they can use it to catch up with high skill workers. LLM use didn't seem to benefit high skill workers as much suggesting that high skilled workers are the ones who're the mostly likely to lose in the adoption of AI.
I'm probably more on the Tyler cowen camp.
I wrote a short note about the potential impact on the labour market. We mostly use our brain powers for job stuff instead of intellectual pursuits. The job market impact is more relevant for most people.
https://substack.com/@mdnadimahmed888222/note/c-131001514?r=o2bbq
Interesting point. I believe you are suggesting that there will be a shift in IQ distribution, as LLMs enable individuals to qualify for jobs that previously required higher IQs (using IQ as a rough proxy for intelligence or ability to do stuff). For example, someone that couldn't be a paralegal before now can, but it seems this breaks down eventually. Will current paralegals now become lawyers?
This situation appears to create a bottleneck in various professions. If, in my example, if the current paralegals can't jump to lawyers, we now have more people that can be paralegals with LLM suppressing the wages of paralegals.
I don't think it will work this way. In my example, the current paralegals will be able to get more work done, and the "lower IQ" people still won't be able to compete. In the end we'll need fewer paralegals. In this case the paralegals will not necessarily use more neurons as they manage the LLM more than writing and thinking about briefs.
We'll see, but the time-use data suggests that if folks have more free time, they aren't going to use it for intellectual pursuits. If presented with the choice of a cookie or broccoli, most will choose the cookie. We have already seen that, given the choice of a book or a screen, most choose the screen.
The price elasticity of demand governs this. If dropping the price of legal advice to a tenth of current prices increases demand for legal services a hundredfold, then we could expect employment to go up. Some of the new employees will necessarily have "lower IQ" than the existing ones.
That level of price elasticity seems plausible for legal services - most people outside of large wealthy corporations use lawyers only rarely. It seems like there is a lot of latent demand priced out of the market.
Of course the lawyers' unions (bar associations?) could frustrate this by colluding to keep prices high.
I'm not sure that AI drops the price of legal services. I still think there is a bottleneck issue. Yes, paralegal work will get cheaper, either by more paralegals getting paid less while using AI or super paralegals that can do more with AI, so we need fewer of them. In the end, all legal work has to be approved by lawyers. Will lawyers use AI to proof the AI work done by the paralegals? That seems risky, at least for now. Even if paralegals can produce more of their work for less, lawyers can't (so far).
I think this goes for various professional activities; as long as someone has to review the work done by AI then prices don't change much.
Predicting the future is a risky endeavor. Either way, AI will be disruptive, but I don't think it is clear yet exactly how.