"AI is here to stay": never backed by any actual evidence

"AI is a bubble": backed by the actual economics playing out in the real world

@eniko True, but I'd like to add: "AI existed before ChatGPT and will keep existing well after OpenAI is relegated to the history books."

@j_bertolotti @eniko

You can still make expert systems with Prolog, but nobody's going to give you $100B to do it.

@petealexharris And that is indeed part of the problem. Increasingly large and power-hungry deep neural networks have sucked all the oxygen from other forms of AI.
@eniko
@j_bertolotti @petealexharris @eniko you should promise shareholders they won't have to pay humans anymore, that seems to motivate them for *some reason*

@j_bertolotti

But the term "AI" was crafted as a bit of petty politicizing, and does not refer to a single technology, making it redundant.

That redundancy is why LLM reference was able to successfully usurp the term.

If we need to relinquish the term along with the bursting of the bubble, we will be no worse off for it.

@eniko

@eniko It **A**in't **I**nevitable. (not sure if the markdown gets rendered)

Last line from Luke Nickle's latest living rent-free in my mind: https://www.youtube.com/watch?v=Z_mzYS-ET8o

There will be some residue when this pops, I don't think the technique will disappear from the face of the earth. But the fact that it's still a fringe opinion to call the most obvious bubble in my lifetime for what it is, is just baffling to me.

F**k Your AI

YouTube

@eniko
Expenditures: >650 bln per year
Revenue: <60 bln per year.

And that revenue includes the 10 bln that Microsoft hands to OpenAI and OpenAI hands back to Microsoft. As @pluralistic said, to call that an accounting trick is doing injustice to the noble accounting trick. This is just ordinary fraud.

This clearly can't go on forever. And as Stein's law has it, anything that can't go on forever eventually stops. The bubble will pop, and given the state of the stock-market it will be ugly.

@sstendahl @eniko @pluralistic According to Ed Zitron, it's 1 Trillion of investment next year. Or $125 for every living human on earth...
@sstendahl @eniko Thanks for making me aware of him. Instant-buy on bandcamp!
https://lukenickled.bandcamp.com/album/singing-in-the-woods-vol-i
Singing in the Woods Vol. I, by Luke Nickle

17 track album

Luke Nickle
@eniko @sstendahl it'll be just like NFT or "blockchain" or dotcom or so on, yeah
still a thing, but slinking in the background, not the main focus
@eniko the logical conclusion of those two premises is: "bubbles are here to stay"

@eniko "But I asked the AI to write a datepicker UI element in BigFramework with lots of special constraints such as leap years and it made in a week a junior's year of work"

And so could a well-trained monkey. Making yet another rewrite of something that has been rewritten and documented ad nauseam and calling libraries specially designed to make such a rewrite as automatic as stamping license plates is no proof of work ability.

@elrohir we used to just get that shit off stack overflow in a couple minutes without emptying a lake's worth of freshwater
@eniko I want to put this toot on a highway billboard
The inevitability message is designed to grab media space and silence any discussion about alternatives. They too are afraid the bubble is going to burst.
@eniko but economics itself is a bullshit science...

@eniko Or perhaps was a bubble, if this is right about its popping:

https://youtu.be/or8butOTUp8

I wonder if it'll be remembered alongside the South Sea Bubble and the Tulip Mania. Maybe the Mechanical Türk and ELIZA too. And potentially even Heaven's Gate or something on that order for the TESCREAL techbro lot.

The AI bubble won’t survive this question | Ed Zitron

YouTube
@eniko the only players who are really backing it are the very rich- not because it’s good for the world but because it will make them richer!
@eniko it's well to be described as a 'mad race to the plateau'.

@eniko Actual evidence: CorridorKey. If you equate “AI” with the subset consisting of large language models, ignoring 60+years of computer science, then sure, there is a bubble and it’s going to pop because LLMs are an unethical environmental disaster of a dead end.

If you look at the entire 60+ years, though it becomes hard to imagine things like machine vision or go players or even green screen removal going away.

https://github.com/nikopueringer/CorridorKey

GitHub - nikopueringer/CorridorKey: Perfect Green Screen Keys

Perfect Green Screen Keys. Contribute to nikopueringer/CorridorKey development by creating an account on GitHub.

GitHub
@MartyFouts when i say AI i mean the things that are labeled AI and not the things that are not labeled AI
@eniko @MartyFouts that's one way to put it 😁

@eniko So do I. Machine vision has been a branch of AI since the 60s. Deep Blue’s Chess program, AlphaGo, and CorridorKey are all examples of machine learning, a branch of AI since the 50s. All of these are labeled AI in the literature.

Whatever you label it, the research field isn’t going away, its researchers are going to continue to call it AI, and it will still produce good results like CorridorKey, long after the LLM bubble bursts.

@eniko The technology won't go away, but the bubble will burst, and most of the companies will go out of business. However, smaller open source models that anybody can download and install on any sufficiently powerful machine will remain. And for the smallest of those models, even a Raspberry Pi can be sufficient.
What will go away is the hype. The frenzy. The attempts to shoehorn AI into everything whether it makes sense or no.
@eniko "the cat is out of the bag" my guy. the cat is running an extortion racket. it is sitting on a leather couch filing its nails saying "gee, that's a nice bag youze got there, would be a damn shame if little old me was to jumps back in there" while you sob and shovel a billion dollars a day into its litter box.
@eniko "this ship has sailed' yes straight into fucking charybdis