Is Scaling Enough for AGI?

This project analyzes whether scaling alone is sufficient to achieve Artificial General Intelligence (AGI).

Final Report

Full written analysis of the scaling vs AGI debate.

Presentation

Slides used for the final presentation.

Supporting Papers (Evidence for Scaling toward AGI)

• Sparks of Artificial General Intelligence (2023)
• Scaling Laws for Neural Language Models (2020)
• Emergent Abilities of Large Language Models (2022)
• Training Compute-Optimal Large Language Models (Chinchilla, 2022)

Opposing Papers (Scaling is NOT enough)

• TruthfulQA (2021)
• The Illusion of Thinking (2025)
• Scaling Is Not Enough (2026)

Key Argument

This project compares two perspectives:

• Pro-scaling: Larger models lead to better performance and emergent capabilities.
• Anti-scaling: Scaling alone cannot achieve true AGI; limitations remain in reasoning, truthfulness, and generalization.

Conclusion

Scaling is a powerful approach that significantly improves model performance, but current evidence suggests it is not sufficient on its own to achieve AGI.

Author: Che-Wei Hsu, Ming-Chun Lee