Epistemic status: Confident in the core distinction. The underlying Coherence Resolution Mode framework is theoretical and could be wrong, but the category error I'm pointing at is real regardless.

------

There's a term that's achieved near-consensus among AI research leaders: superintelligence. Hinton warns about it. Legg says it's mathematically inevitable. Yudkowsky built a career on it. Bostrom wrote the book.

The term does no coherent explanatory work. What these researchers actually describe, when you look at what they're pointing at rather than the word they use, is super-equipped intelligence: systems with the same cognitive architecture operating with better resources and faster execution.

This isn't semantic pedantry. It changes what we should worry about and what we can do about it.

What They Actually Claim

What precisely do leading researchers say when they invoke superintelligence. The claims cluster into a few categories:

The qualitative superiority claim. Hinton suggests future systems will think thoughts humans could never conceive. Bostrom defines superintelligence as intellect that "greatly exceeds human cognitive performance across virtually all domains." The implication is a difference in kind, not degree. It's a form of understanding fundamentally inaccessible to human intelligence.

The architectural transcendence claim. Shane Legg points to hardware: biological neurons fire at 100-200 Hz whilst silicon operates at billions of Hz. Neural signals propagate at 30 m/s; electronic signals at light speed. The brain runs on 20 watts; AI data centres on hundreds of megawatts. When asked whether human intelligence represents any kind of natural ceiling, Legg's response: "Absolutely not." Hinton thinks "it's quite conceivable that humanity is just a passing phase in the evolution of intelligence."

The recursive explosion claim. Yudkowsky's FOOM scenario: once a system can improve its own intelligence, each improvement enables further improvements, creating an exponential cascade resulting in intelligence "vastly beyond human comprehension."

The dominance claim. Altman describes building systems whose "intellectual capacity will vastly outstrip human potential." Amodei predicts systems "better than almost all humans at almost everything, and eventually better than all humans at everything."

These aren't claims about faster computation or larger models. They ...

Posts

Wikitag Contributions

Comments

Posts

Wikitag Contributions

Comments