Greg Kamradt: AGI's 'general' means doing what it wasn't trained for, via self-built tools
GregKamradt · x · 2026-09-09
Greg Kamradt argues that the 'general' in AGI means a system can do things it wasn't trained to do. If a model's weights are frozen (he assumes Google's astra's still are), any capability gain must come from external memory and tools the model builds for itself. He cares less about what a model can one-shot and more about what it can accomplish through building its own tools given enough time and inference.
Related event: Debating AGI: Untrained Capability vs Ability to Learn Anything(5 posts)→
More from AGI Musings
- Nikita Bier: AI-generated renovation plans are basically construction-ready now — aarthir · 2026-09-09
- Report: coding agents boost code output but gains shrink sharply before production — omarsar0 · 2026-09-09
- The AI goalpost keeps moving: from can't multiply to IMO gold, still not enough — IgorCarron · 2026-09-09
- The future is unimaginable: AI can't contemplate, so good goals must be discovered by humans — sebkrier · 2026-09-09
- Local Bad Behavior In RL Doesn't Produce Emergent Misalignment, Per Hacker-Opus — 1a3orn · 2026-09-09
- A 10-Minute Method to Calculate Your 'AI Exposure' — and Why the Unmarked Tasks Matter Most — ravikantagrawal · 2026-09-09