Hey folks, my name is Kevin. I've worked in software for a lot of years. I've co/founded several startups and my most recent role was Head of AI at a consulting firm that was acquired by Accenture last year. I left that role last summer and I've been working on a Masters in AI through UT Austin since then (graduating in Dec!). I'm currently working on a research project through one of their robotics labs focused on World Models. I also do some independent consulting and I'm assisting a local startup working on text -> video generation of sign languages. I'm super passionate about AI and always up for nerdy chats 🙂

Welcome! Those sound like some awesome projects.

Thanks! Yah world models are wild and generating sign videos is a surprisingly deep and under-explored area of AI.

Do happen to have any recs for reading up on the latest with world models? They've come up quite a bit in past #C06AAG8LU9M discussions. But too new for any books yet 😂 (we were thinking of maybe doing papers too)

Yah, WM are super cool and there are lots of angles to the topic. The original World Models paper is a great starting point. Dreamerv3 is a popular architecture and they discuss training RL agents within the model. The Genie 2 paper is a great deep dive, especially their architecture for unsupervised learning of actions from raw video frames. Lastly the recent Founder paper (July 25) is a bit technical but it's a really creative example of shaping rewards for RL agents inside the rollouts of a world model.

Awesome, thank you!