In the meantime keep learning and practicing cs fundamentals, ignore hype and build something interesting.
I don't really agree with the reasoning [1], and I don't think we can expect this same rate of progress indefinitely, but I do understand the concern.
If software falls, everything falls.
But as we've seen, these models can't do the job themselves. They're best thought of as an exoskeleton that requires a pilot. They make mistakes, and those mistakes multiply into a mess if a human isn't around. They don't get the big picture, and it's not clear they ever will with the current models and techniques.
The only field that has truly been disrupted is graphics design and art. The image and video models are sublime and truly deliver 10,000x speed, cost, and talent reductions.
This is probably for three reasons:
1. There's so much straightforward training data
2. The laws of optics and structure seem correspondingly easier than the rules governing intelligence. Simple animals evolved vision hundreds of millions of years ago, and we have all the math and algorithmic implementations already. Not so, for intelligence.
3. Mistakes don't multiply. You can brush up the canvas easily and deliver the job as a smaller work than, say, a 100k LOC program with failure modes.
ETA:
You updated your post and I think I agree with most of what you said after you updated.
All relevant and recent evidence points to logarithmic improvement, not the exponential we were told (promised) in the beginning.
We're likely waiting at this point for another breakthrough on the level of the attention paper. That could be next year, it could be 5-10 years from now, it could be 50 years from now. There's no point in prediction.
People like to assume that progress is this steady upward line, but I think it's more like a staircase. Someone comes up with something cool, there's a lot of amazing progress in the short-to-mid term, and then things kind of level out. I mean, hell, this isn't even the first time that this has happened with AI [1].
The newer AI models are pretty cool but I think we're getting into the "leveling out" phase of it.
I don’t think that follows at all. Robotics is notably much, much, much harder than AI/ML. You can replace programmers without robotics. You can’t replace trades without them.
Are you so sure?
Almost every animal has solved locomotion, some even with incredibly primitive brains. Evolution knocked this out of the park hundreds of millions of years ago.
Drosophila can do it, and we've mapped their brains.
Only a few animals have solved reasoning.
I'm sure the robotics videos I've seen lately have been cherry picked, but the results are nothing short of astounding. And there are now hundreds of billions of dollars being poured into solving it.
I'd wager humans stumble across something evolution had a cake walk with before they stumble across the thing that's only happened once in the known universe.
Your exponential problems have exponential problems. Scaling this system is factorially hard.
https://en.m.wikipedia.org/wiki/Moravec%27s_paradox
https://harimus.github.io/2024/05/31/motortask.html
Edit: just to specifically address your argument, doing something evolution has optimized for hundreds of millions of years is much harder than something evolution “came up with” very recently (abstract thought).
You've got this backwards.
If evolution stumbled upon locomotion early -- and several times independently through convergent evolution --, that means it's an easy problem, relatively speaking.
We've come up with math and heuristics for robotics (just like vision and optics). We're turning up completely empty for intelligence.
>Only a few animals have solved reasoning.
the assumption here seems to be that reasoning will be able to do what evolution did hundreds of millions of years ago (with billions of years of work being put into that doing) much easier than evolution did for.. some reason that is not exactly expressed?
logically also I should note that given the premises laid out by the first quoted paragraph the second quoted paragraph should not be "only a few animals have solved reasoning" it should be "evolution has only solved reasoning a few times"
Any citations for this pretty strong assertion? And please don't reply with "oh you can just tell by feel".
We also should end the exploitative nature of globalization. Outsourced work should be held to the same standards as laborers in modern countries (preferably EU, rather than American, standards).