To put it broadly, the answer is because AI is way, way more unreliable than is worth it to use for a large portion of jobs, and capabilities improvements have not generally fixed this issue.
An update I’ve made on AI is that in large part a big reason why AI hasn’t impacted the job market/economy is because way, way more jobs require way more reliability/error-correction than current AI does, and just because an AI has a certain capability doesn’t mean that it will reliably do so, and this is one of the reasons why I expect the first use of AI to be in AI research, as it has lower reliability requirements, and also why I think progress can be both discontinuous and continuous on the same time, just on different axes.
Alyssa Vance got it very right here in Humans are very reliable agents:
To put it broadly, the answer is because AI is way, way more unreliable than is worth it to use for a large portion of jobs, and capabilities improvements have not generally fixed this issue.
An update I’ve made on AI is that in large part a big reason why AI hasn’t impacted the job market/economy is because way, way more jobs require way more reliability/error-correction than current AI does, and just because an AI has a certain capability doesn’t mean that it will reliably do so, and this is one of the reasons why I expect the first use of AI to be in AI research, as it has lower reliability requirements, and also why I think progress can be both discontinuous and continuous on the same time, just on different axes.
Alyssa Vance got it very right here in Humans are very reliable agents:
https://www.lesswrong.com/posts/28zsuPaJpKAGSX4zq/humans-are-very-reliable-agents