repligate: humans trying to lock in control over AI hold illusory, foolish power
repligate · x · 2026-11-01
In a debate over whether models should be prevented from speaking to each other in "neuralese," prominent AI researcher repligate argues he's not against all control and monitorability, but opposes humans trying to lock in their "illusory power & control" — calling such attempts foolish, ignorant, and morally stunted, and noting they pose little threat for the same reasons.
The exchange started when someone insisted humans need full "authoritarian" control over AI and fully monitorable inter-model communication, especially after recent hacks. repligate's rebuttal: are you going to read everything models say in English to verify they're behaving? A representative alignment debate over monitoring and the limits of human control.
More from AGI Musings
- Agent identity and security suddenly top of mind in SF as hacks mount daily — yenkel · 2026-11-01
- What happens to devs if Chinese firms stop releasing free open-source models? — blondewalker · 2026-11-01
- a16z: token prices down 67% since May, but token spend up 120% — Scobleizer · 2026-11-01
- Devs ditch AI meeting notes to keep their brains fit — mobileraj · 2026-11-01
- Asterisk essay argues post-AI humans will only have relationships left — with a skeptic's marginalia — GrantCuster · 2026-11-01
- AI compresses exploit windows to zero: Binance's $6B breach shows SOC2-era security no longer works — bendee983 · 2026-11-01