121. Alexei Baevski - data2vec and the future of multimodal learning

<p>If the name data2vec sounds familiar, that’s probably because it made <a href="https://ai.facebook.com/blog/the-first-high-performance-self-supervised-algorithm-that-works-for-speech-vision-and-text/" rel="noopener" target="_blank">quite a splash</a> on social and even traditional media when it c