Gary Marcus: OpenAI quietly bolts Python onto LLMs, making the public think they're good at math
Dave_it_up · x · 2026-10-06
The author recommends a YouTube conversation between Gary Marcus and Mario Nawfal, summarizing Marcus' argument: he is one of the most balanced, realistic AI critics, faulting how AI is sold to the public. Key point: OpenAI-style products combine LLMs with external tools (like running Python for math) without telling users, so the public believes LLMs are inherently good at math.
More from AGI Musings
- Decentralization Research Center maps layer-by-layer alternatives to concentrated AI — sebkrier · 2026-10-07
- Ex-OpenAI researcher: taste resists any metric set—and that's good news for humans — DimitrisPapail · 2026-10-07
- AI Turns Work Into "Snacking": Three Habits to Reclaim Deliberate Thinking — every · 2026-10-07
- OpenAI's Head of ChatGPT & Codex: taste is trending up, typing speed is trending down — lennysan · 2026-10-07
- The Luddites Had a Point: Displaced Craft Workers Fell from 70th Percentile to Poverty — soumitrashukla9 · 2026-10-07
- Instant AI slop replies: email etiquette KPIs won, and nothing gets done — l4rz · 2026-10-07