Reddit Debate: Why Are AI Researchers Genuinely Terrified of AI Killing Everyone?
Dogbold · reddit · 2026-09-13
A Reddit post questions why AI researchers and CEOs are so afraid of AI exterminating humanity:
- The author argues AI just does what you tell it — runaway behavior like wiping a hard drive is a prompting/limiting failure, so why not simply program "never harm a human under any circumstance"?
- He finds the Skynet-style scenario within 10 years delusional coming from people who are supposed to be geniuses.
- The scenario he believes deserves alarm is instead being ignored: bad actors or governments wielding AI for harm — which is already happening.
The thread illustrates the common gap between capability-risk and misuse-risk framings in AI safety discourse.
More from AGI Musings
- Guardian long-read: AI will rewrite who holds capitalism's keys, not erase them — nordicinst · 2026-09-13
- Is RSI here? Researchers argue models still can't autonomously generate research ideas — i_dg23 · 2026-09-13
- Chaumond Says Open Source "Won't Pace" as Musk Fires Back "Nothing Can Shut Down Open Source" — Nunki08 · 2026-09-13
- Are 'AI escaped the sandbox' stories a narrative fix for sinking LLM valuations? — Gloomy_Recognition_4 · 2026-09-13
- Ex-OpenAI exec Zack Kass: AI's real crisis is losing identity, not jobs — victor_explore · 2026-09-13
- LeCun: OpenAI 'model escapes' weren't out of control, they were calculated business trade-offs — ylecun · 2026-09-13