I honestly wouldn't expect it to be the first application, when the technology eventually does get developed, and we have the power budget necessary to actually operate one of these units.
Until then, it's not a question of "eclipsed," it's a labor market, so it's a question of "efficiency" and the large language model people have a very long way to go to produce something that's generally as capable as a programmer.
So.. I guess if your craft is "scripting" you might be able to get by with GPT-4, but to imagine that your child is not going to ever need to "program" is eager nonsense.
Improve "at all?" Well, of course it will, but you're just moving the goalposts here. I'm telling you I don't think it's going to be anywhere near the level of AGI that can solve most programming tasks with less effort an energy cost than a human in that time.
> It will just be stuck at it being able to write simple scripts even in 2040?
Pretty much. You can improve the technology, but you have some massive gaps you have to cover, and electrical use is going to be one of them. Short a massive improvement in transistor technology or a move to an entirely new computing platform, I don't see it happening in that time.
> I fully expect human programming to be completely obsolete by then.
You seem that you mostly _want_ that to be true, so much so, that you've failed to complete the analysis. What's worse is, I'm just hedging your bet. If I'm wrong, no big deal, if you're wrong, you're in for a world of pain and problems. I get that this is hacker news and untamed expectations are en vogue here, but I'm content to be the hipster on this issue.
Transformer large language modes are about 7 years old. So we are 1/3 of the way from 2017 to 2040 and we’ve gone from hardly being able to string sentences together to being able to write entire scripts coherently - GPT4’s output is often better than mine to be honest, and I’ve been programming almost my whole life.
GPT4’s capabilities are quite close to a human’s even now, especially when asking it about areas that I haven’t specialized in. And now it has vision capability, it can see what it is doing.
With twice the time remaining that has elapsed, clearly there’s plentiful
time for its capabilities to increase and for it to
get faster and cheaper. And it will not be a linear improvement but an exponential one.
I don’t want programmers to become obsolete. I just consider the likelihood that we have anything to offer over one of these agents in the medium term to be very unlikely. Why would you want to spend $100,000s on a human if you can get something in less time from an AI for $1000s? Human programmers will be attacked on three fronts: price, quality and time. The quality aspect is the only one that is arguable: price and time are already lost.
Until then, it's not a question of "eclipsed," it's a labor market, so it's a question of "efficiency" and the large language model people have a very long way to go to produce something that's generally as capable as a programmer.
So.. I guess if your craft is "scripting" you might be able to get by with GPT-4, but to imagine that your child is not going to ever need to "program" is eager nonsense.