AI doesn’t have to be a black box. We’re a team of physicists, neuroscientists, and ML engineers who are taking a different approach to understanding the fundamental science of artificial intelligence. We’ve already demonstrated that neural networks learn to perform belief updates over world models during inference. Now we’re working on new tools to discover these beliefs in LLMs, decomposing AI models, and bringing light to the black box. Ultimately we believe we can solve interpretability and the nature of intelligence, and provide useful and valuable tools for steering us in a more positive direction for humanity.

Simplex

Org chart

Teams

Offices