[FRIAM] Fwd: The coming AI crash-worse than the Dot Com stock collapse?

[Prof David West]

Marcus and Jon are not incorrect. I do see a problem that they do not, the fact that the vast majority of users/adopters of AI are dramatically less technologically compentent than either of these gentlemen. The manager that is positive that AI will eliminate most if not all of his human employees, the student using an LLM to "cheat," the social media addicts taken with the latest AI fad bots, etc. etc. almost certainly will become disillusioned and turn away from AI.

Perhaps more importantly, all the capitalists who see immediate—not long term—return on investment are not going to remain invested. Lot's of other peoples money will be lost as a byproduct.

The market of Jon Marcuses is not large enough to sustain all the current investment. Maybe one large AI company will survive (ala Amazon that lost tons of money for a long long time).

davew


On Sun, Nov 16, 2025, at 3:51 PM, Jon Zingale wrote:
> I mostly agree with Marcus' sentiment. The dot com analogy may be apt, but it also smells too easy an analog. I find the K-shaped AI adoption to be bizarre. Personally, I do not believe LLMs, nor any particular architecture, to be the be-all-end-all. I suspect we will see a transition away from throwing money at developing the most general form and a move toward more idiosyncratic instantiations. For instance, I continue to think that Deepmind did meaningful work going the RL path with AlphaGo/Atari games and it has yet to come to my attention what happens when Transformers attempt to replicate these successes. Almost every LLM I have met is really really bad at go. This said, AI in their current form, and from this perspective, has been here for a decade. Some have adopted it and use it to surprising effect, others treat LLMs as nothing more than a robust database querying language. What people do with it and how they perceive it will undoubtedly have an impact. In the meantime, I am excited to see what happens as programmers learn to use formal type theories as pidgins and LLMs become more amenable to compositionality.
> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom https://bit.ly/virtualfriam
> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIC http://friam-comic.blogspot.com/
> archives:  5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
>   1/2003 thru 6/2021  http://friam.383.s1.nabble.com/
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20251117/bfaa64c9/attachment.html>