The Paradox of Contemporary AI @3QD

I’ve got another article at 3 Quarks Daily:

The Paradox of Contemporary AI: Engineering Success and Institutional Failure

The title pretty much encapsulates my argument. Yes, LLMs have been a brilliant technical success. But they’ve also led to an intellectual monoculture that narrows the range of ideas on which to draw in moving forward. That, I argue, will make progress much more difficult. I’ve published a longer and more narrowly circumscribed version of that argument on New Savanna, The Shock and the Narrowing: How ChatGPT's Success May Have Compromised AI's Future.

Part of the institutional failure lies in the educational system. Here’s a short note I posted to Facebook that points at that:

I entered Johns Hopkins as a freshman in the fall of 1965. JHU had just gotten Ford Foundation money to establish the interdisciplinary Humanities Center. “Interdisciplinary” was the next Big Thing. In the fall of ‘66 the Humanities Center hosted the (in)famous structuralism conference. And we all know what happened after that. Theory plus Harold ‘Effin’ Bloom.

Since then I believe JHU has acquired two more interdisciplinary humanities centers, but the place is still run by the traditional disciplines. That’s happened all over the place. Interdisciplinary centers and studies programs all over, but it’s all epicycles upon epicycles circling the same old 19th century German categories.

For all that time I’ve heard that too many people are publishing the same damn articles time and again. That problem persists. & I’ve heard that we need to do more to encourage Real Innovation, we always need more of that. But the institutions remain fundamentally the same.

Is there any chance that this is ever going to change? What are the chances that AI will force/allow fundamental change? Or do we have to somehow start anew from the ground up?

AI experts don’t realize how shallow their knowledge is, and neither does anyone else.