i believe it to be easier logistically for AI to figure out some accounting or lawyering than it would be for a robot to be a plumber/electrician/car mechanic, whatever.
circling back on this now, lots of talk in here has been about unskilled labor - but you get a lot more on your return replacing higher skilled labor, because it is expensive.
eg. what's your return on a machine that flips a burger? there's mechanical simplicity here (in reality mcdonald's uses clam grills ever since i worked there at 16) - but the labor you save is only so much. meanwhile...
How much does a Anesthesiologist make in the United States? The average Anesthesiologist salary in the United States is
$414,800 as of December 27, 2022, but the range typically falls between
$359,300 and
$470,200.
so how about we instead just feed some sensors/data points in to a machine, connect it to the actions it can execute based on those inputs, then through software learning of 1,000,000 medical procedures learn how to make the best choice at any given time and let it evaluate those choices constantly in real-time?
you probably aren't going to go from 10 anesthesiologists on staff to 0, but 10 to 2 - why not? and for sure this is easier to do than replace or augment a plumber.
That'll probably happen, but I think it's a very very long way off for a couple reasons. One, companies don't generally adopt new technology quickly. Two, I think it'll take longer than one would expect because there will be a lot of governmental/economic ramifications from the replacement of unskilled labor, which will delay further advancement of AI
even if you have a human making a final decision "eg. this is cancer" the machine is going to help and replaced skilled labor - so even if AI bot can't practice law, it can replace half your paralegals... and this stuff is going to grow and mature crazy quickly because the bounds to it learning are not
that high. eg. the amount of data chat gpt was trained on, and we've just been seeing version 3:
this just comes down to how much time/money/hardware can be thrown at it - but it's not like you can't continue to throw even more of any of these things at it, or that it isn't already going to very quickly eclipse our monkey brains.
"companies don't generally adopt new technology quickly"
these companies are just going to get ran over - or just fully disrupted. slowmo enterprise co. will just have it's market share eaten away and destroyed by some upstart that can move quick and not have legacy processes and systems and etc etc. holding it back. some company you have never heard of or hasn't been born yet will arise and just eat the lunch of some enterprise because they can do the same thing, better, with 10% of the staff. this does have the caveat of them being able to get data somehow - the big plus an enterprise has going for it is 1. big budget today and 2. all of their data. eg. i have tools available to me to make a great loan prediction model - i actually used machine learning to invest in loans on lendingclub and prosper.com starting in 2018 - but i also don't have the history of millions of loans like chase bank does. (well kinda, lending club stopped even offering loans to the retail market - but back when i made this, you could download a decent amount of info about the millions of their own loans)
