Seonglae ChoArtificial General Intelligence
Since AGI is based on internet data, it should be compared to humanity's capabilities as a whole rather than individual human abilities. Therefore, the point of reaching AGI begins from the moment it starts helping humanity, and the point of reaching ASI (Superintelligence) is when AI's contribution to human progress exceeds that of humans.
The very existence of humans is proof that AGI is possible. In other words, it demonstrates that such generalized learning is physically and efficiently achievable. This is possible because humans use DNA as a prior, with evidence showing highly sample-efficient algorithms refined through the bottleneck of DNA as a minimal MDL (minimum description length) of distilled information.
AGI Notion
Carbon-based intelligence is merely a catalyst for silicon-based intelligence
It’s coming. It’s inevitable
Planning beyond
Checklist to AGI
If controlling AGI is fundamentally impossible, is suppressing development the only way forward?
Ilya Sutskever 2025
AGI is intelligence that can learn to do anything. The deployment of AGI has gradualism as an inherent component of any plan. This is because the way the future approaches typically isn't accounted for in predictions, which don't consider gradualism. The difference lies in what to release first.
The term AGI itself was born as a reaction to past criticisms of narrow AI. It was needed to describe the final state of AI. Pre-training is the keyword for new generalization and had a strong influence. The fact that RL is currently task-specific is part of the process of erasing this imprint of generality. First of all, humans don't memorize all information like pre-training does. Rather, they are intelligence that is well optimized for Continual Learning by adapting to anything and managing the Complexity-Robustness Tradoff.
