Gary Marcus Fires Back: 'Deep Learning Hitting a Wall' Was Right About Tools and Harnesses
GaryMarcus · x · 2026-09-08
Gary Marcus pushed back against critics who claim he repeatedly predicted walls at o3, GPT-4, etc. He says his 2022 'deep learning is hitting a wall' argument was that the field would eventually need to complement deep learning with symbolic mechanisms — and that's exactly what happened with loops, code interpreters, harnesses, and tools. 'My paper was fully correct. The issue is shitpoasters' inattention to detail.'
More from AGI Musings
- OpenAI chief economist: AI created ~1M US jobs vs ~200K AI-attributed layoffs — robseamans · 2026-09-08
- OdiseIA on the artist who 'married an algorithm': love, loneliness and tech — vrbenjamins · 2026-09-08
- Denny Zhou revisits his 2022 'teach LLMs to reason' talk's closing slide — denny_zhou · 2026-09-08
- Self-Improving AI's First Payoff Is Cheaper, Not Smarter: Ben Lorica on the Harness Advantage — bigdata · 2026-09-08
- AI could 1000x knowledge work output, pushing universal income to $5M+ buying power — davidpattersonx · 2026-09-08
- OpenAI Mathematicians: Model Reasoning Traces Look Strikingly Like Expert Humans' — a16z · 2026-09-08