Peak LLM debate: VCs says models won't get smarter, critics say you can't tell anymore
repligate · x · 2026-11-19
@jdnoc argues LLMs won't get any "smarter" — only more efficient, more capable via plugins/memory/context/skills, and more specialized, comparing current iterations to the iPhone's 10th-generation plateau. Repligate boosted the counterpoint: if you can't tell AIs are still getting smarter, it's probably because they're now smarter than you and you can no longer assess their intelligence.
Related event: Debate Erupts Over Whether LLMs Have Peaked in Capability(2 posts)→
More from AGI Musings
- Viral 'AI-written anti-AI song' video isn't AI thinking, just parroting with visuals — TinfoilTricorn · 2026-11-19
- Thomas Nagel's AI consciousness interview draws fire for under-argued assertions — rgblong · 2026-11-19
- 'Consciousness License' meme mocks everyone piling into the AI consciousness debate — alexisgallagher · 2026-11-19
- Katja Grace: most people don't viscerally grasp that AI soon risks their family's lives — KatjaGrace · 2026-11-19
- Catching an AI's error doesn't prove you're smarter — it just exposes a specific weakness — repligate · 2026-11-19
- Corollary: ASI will be largely invisible to the general public, misattributed even when noticed — repligate · 2026-11-19