Nicholas Thompson: AI extinction talk is overblown, but alignment research still deserves priority
GaryMarcus · x · 2026-09-17
Nicholas Thompson argues it's fair to believe discussion of AI extinction risks is overblown, while still hoping this moment drives greater prioritization of alignment, interpretability, and risk-reduction work across labs, government, and non-profits.
More from AGI Musings
- Researchers debate AI risk framing: instant-apocalypse narratives vs. boiling-frog gradual harm — eigenhector · 2026-09-17
- Is domain-specific training a constant-factor gain or a scaling exponent change? — eigenron · 2026-09-17
- DeepSeek engineer's essay: AI will outdo me in a year, but I'll keep coding — enginetown · 2026-09-17
- Reddit essay: redefining agents via persistence substrate vs execution substrate — D777Castle · 2026-09-17
- AI safety and 'normal technology' camps converge on policy despite worldview gaps — dhadfieldmenell · 2026-09-17
- Oracle India cuts 3,000 jobs, Adidas axes half its India tech hub staff — DrDatta_AIIMS · 2026-09-17