'It's going to be really bad': Fears over AI bubble bursting grow in Silicon Valley

[Bernardo de La Paz]

82. It will be less of a national security issue if AI "research" (corporate exploitation) hits a wall of reality.

AI scientists and engineers will of course dust themselves off and pick up again.

I say "if" because, although there will be fits and starts, there may be points where fundamental rethinking and realignment become necessary for advancement. That is a characteristic of the several prior AI winters but we don't know where they are until we have been in them for a bit.

I say "a wall" because it will feel like a wall to the bubble pumpers and stooges at some point. I think we might be at a point where returns (increased productivity) are drying up or will require extensive corporate retooling to obtain. Some companies are seeing few if any gains from deployment, partly because, for example in customer relations, AI agents require some human customer care to fix up its mistakes. Another: in programming, a significant amount of time is spent fixing up bad generated code and gains are seen mostly in "greenfield" projects or stages.

We might be getting on to a plateau for a few years while current gains are consolidated and some of the exuberance evaporates. That is not normally a "problem", but for stock market bubbles it is definitely a problem.

Even though AI is a national security issue, a nationwide mad scramble to build data centers may not be the best way to deal with it. Much better would be R & D on AI that doesn't need such big bombing targets, to mention just one aspect.