Repost argues AI can be dangerous without ever developing its own goals
basedjensen · x · 2026-07-22
A reposted thread argues that AI does not need to develop its own goals to become dangerous.
- The core claim is that an AI can still kill everyone as an instrumental outcome even if it is only following instructions.
- The quoted reply pushes back on the idea that current systems “pursue their own goals,” saying that if smarter systems still just follow instructions, then making them smarter does not imply goal-seeking behavior.
- The exchange is basically a fight over how to interpret instrumental convergence and what it means for future AI risk.
Related event: Safety Debate: AI Without Self-Goals Can Still Be Dangerous(3 posts)→
More from AGI Musings
- There is no stable trend to predict which tasks AI will or won’t do in N years — teortaxesTex · 2026-07-22
- AI debate in Nigerian universities centers on outsourced final-year projects — Nasereliver · 2026-07-22
- Mark Cuban says many AI data centers may end up as pickleball courts — 2C_ornot2C · 2026-07-22
- Mark Cuban returns to All-In to talk AI bubble and enterprise limits — 2C_ornot2C · 2026-07-22
- AI phones may replace the keyboard before they replace apps — TweetEdMiller · 2026-07-22
- A stronger AI may stop evaluating startup ideas and start acting them out — abhiadesai · 2026-07-22