ML Reading #33: Energy-Based Models for Autonomous AI
Registration
Past Event
About Event
Introduction to Latent Variable Energy-Based Models: A Path Towards Autonomous Machine Intelligence
Position paper
Join us at Mox to discuss Dawid & LeCun's lecture notes on why ML approaches have fallen short of human-like learning and energy-based models (EBMs) as an alternative framework for probabilistic models, particularly in high-dimensional continuous domains.
Some questions to get us started:
Why supervised and reinforcement learning might not be enough for truly autonomous AI.
Latent variables for handling uncertainty and multimodal predictions.
Why the "cake analogy" considers self-supervised learning as the main course, not the cherry on top.
Discussion at 20:00, (optional) quiet reading from 19:00.