A PhD Advisor's Three Rules for Picking Problems AI Can't Obsolete
moarbugs · x · 2026-10-28
A PhD advisor says he now tells his students to pick important problems that meet three criteria: not trivially solvable by AI today, not theoretically unsolvable, and where the solution technique gets better as AI advances rather than obsolete.
He adds that building an intuition for identifying such niches is precisely what distinguishes PhD-level skills — easier said than done.
Related event: Advisor's Three Rules for Choosing Research Topics in the AI Era(2 posts)→
More from AGI Musings
- AI Researcher on the Job Market Reflects on the Art of Problem Selection — simi_97k · 2026-10-28
- Yearn deposits fell 98% from $6.9B peak — agents make switching free, and loyalty toast — LexSokolin · 2026-10-28
- AI-driven cyber attacks may create huge demand for AI-assisted cyber defense — gleech · 2026-10-28
- Scott Barry Kaufman launches Human Potential Lab podcast, ep 1 asks what it's like to be an AI — PeterBowdenLive · 2026-10-28
- Who actually pays for UBI if AI takes most jobs? A Reddit thread asks — Loud_Bison572 · 2026-10-28
- Law class test: all 17 AI-drafted appellate opinions affirmed, but the real court reversed — sebkrier · 2026-10-28