I think it'll be a while (10-20 years?) until we have AGI good enough to at least potentially replace a non-trivial number of workers as opposed to just being a

I think it'll be a while (10-20 years?) until we have AGI good enough to at least potentially replace a non-trivial number of workers as opposed to just being a productivity tool. LLMs are clearly just a tool. We will of course eventually create human-level AGI, but things like emotions and empathy will come later, if at all, which may limit its utility in interacting with humans and therefore usefulness for many jobs. The major job categories that AGI might be capable of is "professional and business" office jobs and many government jobs, which each account for about 15% of US jobs, but what that leaves is the other 70% which includes not only "blue collar" jobs such as trades, manufacturing, construction, transportation, but also major employers such as healthcare, education, leisure & hospitality, etc. The future doesn't always turn out as expected though, especially when it comes to sci-fi visions of people in flying cars, living on mars, etc. AGI and robotics may well turn out to be the same - we may imagine humanoid robots and human-like "remote workers" doing many jobs, but it'll probably not be like that. Humanoid workers are probably about as likely as flying cars - problematic and just not that useful. AGI "workers" may well be just agents that are given well defined tasks to execute as opposed to fully autonomous human worker replacements.