Can AI play hide-and-seek?

Let's kick off with something called Reinforcement Learning. Here a cool demo from OpenAI back in 2019. It's a multi-agent interaction on a simulated hide-and-seek environment.

Full article https://openai.com/index/emergent-tool-use/

https://youtu.be/kopoLzvh5jY?feature=shared

Did you know OpenAI has a Superalignment Team that works on teaching powerful AI to be supervised by less powerful models? Fun fact both co-led of the team Ilya Sutskever and Jan Leike has left OpenAI. Could it be that Sam wanted it to be uncontrolled😎?

AI word of the day✍️

AI Alignment: aims to make AI systems behave in line with human intentions and values.

Read more here: https://openai.com/superalignment/

ChatGPT got more rizz than @beka_cru and basically y'all

Kevin and Nando are notable figures in AI, specializing in Bayesian machine learning, neural networks, RL, and more. They have praised this book highly. Although I haven't read it yet, I trust it's excellent.

😎 I've met Kevin once, in-person, and he's such an amazing guy. He's also the author of the famous book series on Probabilistic Machine Learning.

What If We Run Out of Data to Train LLMs? 🤔

Ever wondered if we could actually run out of data to train our massive language models? 📉

Tomorrow, I'll be hosting Niklas Muennighoff(Research Engineer at Contextual AI) to present an amazing paper called Scaling Data-Constrained Language Models

Add this to your calendar: https://cohere.com/events/niklas-muennighoff-2024-06