Domingos reignites AI values debate: control machines via objective functions, fear "artificial stupidity"
AlexTensor · x · 2026-11-18
In a debate sparked by whether teaching ethics to statistical software makes sense, Pedro Domingos restated his camp's position: people like Neel believe machines should hold values and be throttled preemptively against extinction risk, while he and Yann LeCun think that's nonsensical — AIs are machines to be controlled via objective functions. The real danger, he argues, is artificial stupidity, and what we need is more intelligence, not less. A lot is riding on this disagreement.
More from AGI Musings
- OpenAI Foundation board member Paul Christiano puts 1-in-3 odds on AI overthrowing government within a year — mark_k · 2026-11-18
- Humanities professor ends homework, moves all learning to in-class close reading to counter AI essays — _akpiper · 2026-11-18
- Coinbase CEO Brian Armstrong: open-weight model teams can beat frontier models, agents need a price-based marketplace — thederbiedone · 2026-11-18
- Talk on computing education in the age of AI at University of Seville — moenig · 2026-11-18
- By 2026, white-collar job applications all route through AI screens that reject before humans ever look — felpix_ · 2026-11-18
- SpaceX CFO: Starship's first second-stage recovery opens path to orbital compute, another 10× cost drop — XFreeze · 2026-11-18