5 hours ago
- The book lacks evidence or discussion of how AI might achieve superintelligence, merely asserting it without argument.
- Authors rely on narrow AI successes (e.g., chess, protein folding) but do not address whether these tasks are analogous to broader intelligent behavior.
- They admit humans do not know what intelligence is and that current AI is shallow, yet claim superintelligence is imminent.
- Their definition of intelligence as 'prediction and steering' is tautological, essentially saying computers are intelligent because they do what computers do.
- They argue we don't need to understand intelligence to build it, deferring to future researchers or AI itself—a circular and unsubstantiated claim.
- The book misrepresents the mystery of AI: we know AI finds correlations, not that it thinks, yet they use ignorance to imply potential superintelligence.