You probably should know what VLAs are. It's the newest generation of models, in heavy use by Robotics & Automated Vehicles. These are vision-language-action models that take the work of LLMs and Multi-Modal data into the realm of real-world actions.
Here's some videos I can recommend watching if you're not ready to dive into coding:
- TWIML Interview a year ago with Physical Intelligence's co-founder about Pi0 - watch here
- A well made history of VLAs by a PhD in Robotics, published 6 months ago - watch here
- An ongoing in-depth series (currently 5 hours!) about VLAs by Agoda's Data Science Director. Covers Pi0, Pi0.5. Latest video on Gr00t released less than 2 weeks ago - watch here
All three of these channels are worth subscribing to!