A Letter from a Machine Learning Engineer
5 hours ago
- An anonymous ML engineer from a frontier AI lab claims that labs like theirs lack a competitive moat and face internal stress, with executives quitting and IPO uncertainty.
- The engineer predicts that open-weights models will match or surpass closed models like Fable 5.1 within a year, and that LLMs are mathematically incapable of achieving AGI (citing a paper based on Cantor's diagonalization).
- The author (Nemin) acknowledges improved AI coding tools but maintains that human oversight is essential; agents serve as helpful assistants but cannot replace programmers entirely.
- The author argues that recent tech layoffs attributed to AI are actually due to hype and market normalization after COVID/ZIRP, not genuine AI capability.
- The engineer warns that online AI discourse is poisoned by paid influencers and that CEO forecasts are unreliable.
- The author finds the email credible but unverifiable, and reflects that even if frontier labs collapse, issues like copyright, environmental harm, and disinformation would remain.