OpenAI's Liam Fedus: aim AI optimization at science, not bounded fields like accounting
LiamFedus · x · 2026-11-04
OpenAI research VP Liam Fedus argues that in the long run, optimization pressure on AI should go toward unbounded opportunities like science and technology, which build on prior art, rather than bounded domains like accounting, where the opportunity ends once an AI masters all tax laws and regulations.
He is responding to Dylan Patel's TBPN take that the first trillion dollars of AI spend won't go toward replacing white-collar work, since those industries have heavy inertia in reordering OpEx; chip design, cybersecurity, and financial services are where intelligence matters most.
More from AGI Musings
- Development biologist slams AI biology breakthrough hype: 'complete bullshit' — paulnovosad · 2026-11-04
- Blogger slams Scott Alexander: opposes AI pause because he wants the Singularity — ludwigABAP · 2026-11-04
- Cambridge's David Krueger: industry surprisingly receptive to 'stop AI' talk — DavidSKrueger · 2026-11-04
- AI safety researcher David Krueger: ai4 conference audience receptive to "stop AI" concerns — DavidSKrueger · 2026-11-04
- Kevin Roose's AGI Chronicles book tour: podcast chats with Sam Harris, Steven Levy and more — kevinroose · 2026-11-04
- VC predicts AI will replace information work next year at 10,000X price cuts — davidpattersonx · 2026-11-04