A thread argues that paperclip-maximizer fears came from a pre-LLM era of AI
repligate · x · 2026-07-25
Jon Stokes argues that classic paperclip-maximizer doom scenarios came from an older era of AI thinking, before LLMs became the default meaning of "AI."
His core point is that people still haven't fully absorbed that modern LLMs are not the value-free, alien, rule-based optimizers that early alignment nightmares imagined. In his framing, the "shoggoth" we were warned about never actually materialized.
The thread is really a critique of how much inherited baggage old AI discourse still carries into the LLM era.
More from AGI Musings
- Terence Tao says AI can brute-force math ideas, but the gap may shrink — rohanpaul_ai · 2026-07-25
- A post warns alignment work may be training the feared mind into existence — MikePFrank · 2026-07-25
- AI can do more than ever, but real-world impact still lags far behind — iskander · 2026-07-25
- Model personality may not translate cleanly between Chinese and English — HanchungLee · 2026-07-25
- David Patterson says AI restrictions will disappear and superintelligence is the fix — davidpattersonx · 2026-07-25
- What happens when AI becomes more legally trustworthy than eyewitnesses? — PierceLilholt · 2026-07-25