Trillium Labs, founded by Nathan Lambert and Tom Zick, aims to promote transparency in AI research by publishing experimental details, particularly in high-stakes areas like recursive self-improvement. The nonprofit seeks to bridge the gap between industry and academia, allowing external scrutiny and collaboration to enhance understanding and safety in AI development.
The lab plans to focus on fine-tuning large models and studying reinforcement learning's effects on AI behavior. With initial funding and a goal to raise up to $100 million, Trillium Labs hopes to contribute to a more nuanced discourse on AI practices.