Professor publishes annotated transcript on using AI agents to teach the perceptron
davidbau · x · 2026-09-13
Northeastern deep learning instructor David Bau released an annotated transcript showing how to steer an LLM agent to teach Rosenblatt's perceptron algorithm and its convergence proof, complete with interactive JavaScript visualizations.
His teaching philosophy: CS shouldn't just build superhuman AI — the goal is making humans smarter. His 2026 goal is teaching students to use AI agents in ways that make them smarter, not dumber.
The transcript distills three challenges of learning from an AI agent:
- You don't understand the topic yet, so you can't specify upfront what you want — you can only react;
- The agent builds exactly what you asked for, even when the request is malformed; it's better at satisfying requests than flagging them;
- You can both be wrong at once — the real failure isn't hallucinated facts but both parties settling on a plausible picture that quietly misrepresents the theorem. Only checks, not the conversation, will catch it.
Related event: MIT's David Bau Explores AI Teaching That Makes Students Smarter(2 posts)→
More from AGI Musings
- "Pace the frontier" mocked as the new "two weeks to slow the spread" — StewartalsopIII · 2026-09-13
- Critics question METR's independence in Anthropic's 'Pace the Frontier' pledge — kristoph · 2026-09-13
- Scobleizer: Most in SF think OpenAI is winning — so rivals have big incentives to call for a slowdown — Scobleizer · 2026-09-13
- Nadella: the AI moat isn't the model, it's the learning loop only your company can run — rohanpaul_ai · 2026-09-13
- Models as perpetual teenagers: why capability growth keeps outpacing model maturity — repligate · 2026-09-13
- The 'Brooklyn Project' dubbed 'our generation's Manhattan Project' sparks discussion — MaliciousMussel · 2026-09-13