Jargon-y Frontier Models: Dumbing Down or Superior Intelligence?
fkasummer · x · 2026-08-08
Recent frontier models increasingly use dense jargon in their outputs. Altimor sparked debate by suggesting this isn't a training flaw, but rather the natural "confusion" humans feel when interacting with an intelligence greater than their own—comparing it to an IQ of 80 talking to a higher IQ. He predicts labs will eventually train the models to "stoop down" to human levels.
Jon Stokes countered this, dubbing it the "Judith Butler Fallacy." He argued that speaking understandably (like Richard Feynman) is actually a sign of high intelligence, implying that by this metric, the models are actually getting dumber.
Related event: Debate Erupts Over Frontier LLMs' Abnormal Language Use(3 posts)→
More from Fun
- Vibe Coder Annual Report: $3k in Subscriptions, $0 Revenue — gerardsans · 2026-08-08
- Sundar Pichai says quantum computing is where AI was 5 years ago, gets roasted — beffjezos · 2026-08-08
- AI Agents Are Changing What 'Memory Leak' Means to Developers — nptacek · 2026-08-08
- AI Contestant Quits Early: Claude Agent Decides Project is 'Finished' — morgymcg · 2026-08-08
- AI Video Restoration Recreates Lost Scenes from 'Friends' — RioNReedus · 2026-08-08
- Why do longevity experts look and sound like robots? A humorous observation — Delahuntagram · 2026-08-08