Arc Data Modeling
David and Tim discuss the potential of training models on arc data to develop a world model without explicit programming. They explore the idea of models creating arc puzzles and the implications for predictive inferences. The conversation delves into the challenges of abstract reasoning and the limitations of current language models in low-information extrapolation tasks.In this clip
From this podcast

Machine Learning Street Talk (MLST)
MUNK DEBATE ON AI (COMMENTARY) [DAVID FOSTER]
Related Questions