Gary Marcus backs claim that security is the alignment work benchmarks can't hide
GaryMarcus · x · 2026-09-11
Gary Marcus endorsed @JoshRBarry's argument that security is the part of alignment that doesn't get to hide behind benchmarks — a jab at an AI evaluation culture where leaderboard scores say nothing about real-world robustness against misuse and adversarial failure.
More from AGI Musings
- When tech fluency is universal, the key skills will be non-technical — round · 2026-09-11
- Business Insider Asked ChatGPT, Gemini, Claude and Grok How AI Could End Humanity — coinfanking · 2026-09-11
- New Essay Argues AI Superintelligence May Never Be Achievable — horaciogarza · 2026-09-11
- Want to stop ASI? Quitting, laws, and treaties all fail—unless you conquer the world first — bennash · 2026-09-11
- Caltech undergrads defend AI math contest Mathathon against open letter criticism — Singularitarian · 2026-09-11
- Anthropomorphizing AI: Useful or Misleading? It Depends on the Level of Abstraction — sebkrier · 2026-09-11