One point I made that didn’t come across: - Scaling the current thing will keep…

AI researcher Ilya Sutskever clarified that scaling existing methods should keep improving models. Sutskever's November 28, 2025 correction rejected the claim that progress would stall, while maintaining that something important would remain missing.

The interview transcript supplies the missing context: Sutskever argued that models still generalize far worse than people, meaning they struggle to apply what they learn reliably beyond the situations used to train them. That position allows continued gains from scale while calling for research into that deeper weakness. The original interview develops that argument; the correction narrows how a summary of it should be understood.