The turbulent AI era is here
11 points by aryeh
11 points by aryeh
AI will either be the greatest equalizer ever invented, or the worst source of injustice.
It's easy to take two opposing extremes like this and state the outcome will be one or the other. Makes for a powerful quote that's hard to prove wrong.
Spot on. He left out "or it will infuse such a proliferation of nonsense and disorder that all out institutions and systems will dissolve into mush", or "it will stabilize an reinforce the existing idiocaracy for a long slow decline," or "it will just stay kind of OK, burning a lot of energy, making some things easier to produce, reducing quality overall, but not really making a huge change". Inconceivable that it will transform everything!
Oh, the guy who thought the Internet was a passing fad is having opinions again.
I think Bill Gates is almost uniquely positioned to offer a perspective on AI worth considering. One statement particularly hit home for me :-
The biggest shift for workers will happen when AI provides nearly error-free work. At that point, it will be able to function on its own without a human checking in on it, and companies will have every economic incentive to let it.
That "when" should be a big "if" instead.
One could say the same for human work.
The difference being that humans already exist, so there's no requirement for us to create new data centres etc to produce them.
And we don't need to justify their existence by being useful enough to produce, because they are sources of value and meaning, not production machines!
I mean, that is true, but I'm not sure what your argument is! Human hands already exist, so there's really no need for us to create tools, etc., yet there are obvious benefits to doing so.
Your argument isn't clear to me either. nemin argued that error-free work provided by AI isn't a given. How does human work related to that?
My statement was that the cost of the current approach to AI is extremely high, and there's no additional cost to humans having jobs, so your suggested human-AI equivalence doesn't seem very robust.
Almost all other tools are inexpensive compared to AI, so that comparison doesn't seem to work either.
I mean, the rate is meaningfully different though, right? Like, on any serious look, AI is about a million times more 'error-y' than a human worker (as eg most humans go their entire career without making the kinds of severe mistakes that AI makes pretty regularly), with no path to get better, no real acknowledgement of the current status of the thing just 'but what if in the future it's good thought, therefore it will be good in the future, therefore it counts as good now'...
There are definitely many things that I would not trust an AI to do today! But also many (and increasing) tasks where AIs outperform humans. (e.g. https://www.science.org/doi/10.1126/science.adz4433)
But if you look at the trajectory of what kind of tasks they can handle, the trend is pretty clear, isn't it? I'm thinking of e.g. METR's time horizon work at https://metr.org/time-horizons
One could, but humans exist now, can be studied, and we can find categories of errors. We can make aids to help (LSP servers, syntax features in programming languages, ASLR) the categories of errors. We don't expect most people to do error-free work, and our cognitive biases prevent us from treating them as demigods. "AI", on the other hand, seems to have some underlying problems that prevent aids from helping. Additionally, human cognitive biases make "The AI said it" into something like a conversation stopping cliche. The bias towards accepting authoritatively-stated answers that appear correct, but contain errors, makes an unpredictably-erroneous AI into something dangerous. If we're going to accept AI answers as definitive or authoritative, as we all have a bias to do, the AI must be nearly error-free.
If that happens, then the entire economy will collapse as nobody will be able to buy anything and then companies will exist exclusively to trade money amongst themselves, which is effectively how the modern economy works anyway. So from my perspective maybe I can stay home and read books instead of getting on Teams calls.
So from my perspective maybe I can stay home and read books instead of getting on Teams calls.
Books are "food for thought" but you'll still need some bread not only read.