What's New Wednesdays: From robot dreams to faster AI iteration
AI is moving beyond models that simply understand the world toward systems that can predict what happens next and act on it.
Join us for a live, demo-driven session connecting one of physical AI’s most exciting research directions with practical tools for faster model development. We’ll explore how World Action Models use future-state prediction for zero-shot adaptation to unseen robot tasks, then show new capabilities for comparing model behavior and keeping research moving.
Whether you’re training models or building AI systems, you’ll leave with a clearer view of what’s next, and workflows you can apply today. You’ll learn how:
Robots can “dream” before they act: Understand how World Action Models (WAMs) jointly predict future world states and robot actions, and why that approach matters for zero-shot generalization in physical AI
To see what is really driving model performance: Use eval tables in W&B Models to compare runs from aggregate scores down to the example-level inputs, outputs, and scores behind changes in performance
To accelerate research workflows: See how CoreWeave ARIA message queueing feature supports a more continuous workflow for follow-up questions and experiment analysis.
We’ll close this session with open Q&A, giving you direct access to our panel of AI experts to workshop your real-world challenges. Save your seat: https://utm.io/usmWy