Mira Murati's Machine Learning Lab Makes a Human-Centered AI Technology Case Built with Customizable Modeling Weights

Think Machines Lab has published a report on building AI that augments human will and judgment. Most AI in use today is trained in a few areas, and then freezes. The report states that this design does not involve the people being used by the model. Instead, researchers at the Thinking Machines lab are looking for AI that is distributed, customized, and shaped by its users.

Proposal for a Thinking Machines Lab

The lab refers to four technical directions. First, it trains robust models through multimodal interaction and customization. Second, create tools that allow people to tune and train themselves to model weights. Third, it develops interfaces that expand the human-machine communication channel. Fourth, it publishes research so that more developers understand how models are made. Together, these directions bring information and alignment closer to users.