I spent some time evaluating the best AI models on interactive block-building tasks. I am surprised by 1) the fragility of these systems when trying to generate creative ideas or update hypotheses, and 2) the vast, but often unnecessary, knowledge and compute they are willing to throw.

Can AI models build a world which today's generative models can only dream of? Presenting BuilderBench (website : t.co/H7wToslhXG). Details below 🧵⬇️

Normalizing Flows (NFs) check all boxes for RL: exact likelihoods (imitation learning), efficient sampling (real-time control), and variational inference (Q-learning)! Yet they are overlooked over more expensive and less flexible contemporaries like diffusion models. Are NFs fundamentally limited?