Multimodal AI trained without paired data
A new training method lets multimodal AI combine vision, audio, and text without ever seeing matched pairs of those signals.
A new training method lets multimodal AI combine vision, audio, and text without ever seeing matched pairs of those signals.