Learning Fast and Slow: Accelerating AI Training with Human-inspired Learning

Current AI systems are trained slowly on a mix of everything—leading to worse performance in languages with less data. This project constructs targeted, human-inspired training interventions to train models faster, and merges models with complementary skills to create models with better reasoning capabilities for smaller languages, while providing causal insights into how these models learn.

Compared to humans, current AI systems learn slow: requiring multiple thousand lifetimes worth of data to train. They also struggle to combine different skills, being worse at the same math problem if asked in a 'smaller' language. To uncouple what AI models can do from the language they work in, we need a modular approach that learns as fast as humans do.