Debate: AI cyber risk is overblown — top models only found basic bugs; bio is scarier
DeepDishEnjoyer · x · 2026-08-31
A short debate on AI risk. One side argues AI cyberattacks aren't that scary: the owner of the most advanced models directed massive compute at two systems and only found basic vulnerabilities, with no critical damage, and everything gets patched. The other replies that bioweapons were limited less by the blueprint and more by the capacity to industrially produce them in a lab.
Related event: AI Risk Debate: Irreversible Catastrophe vs Thermostat-Style Correction(10 posts)→
More from AGI Musings
- Risk of Sudden Higher Ed Collapse as Parents Refuse to Fund Cheating — jfischoff · 2026-08-31
- Rebutting the View That AI Progress Is Just Marketing Hype — joshua_saxe · 2026-08-31
- Census Data: AI Has Minimal Impact on Employment So Far — kuza55 · 2026-08-31
- Why human analogies for AI are wrong — vishalmisra · 2026-08-31
- Census data: most firms aren't changing employment due to AI, hires and cuts split evenly — Afinetheorem · 2026-08-31
- AnjneyMidha: The unprecedented speed and scale of ML systems is wildly interesting — deanwball · 2026-08-31