Testing throwbacks: Meta CEO Mark Zuckerberg tested an old video of him playing a guitar for their new AI model, which in turn has predicted missing parts of a real video.

Taking to Instagram, the billionaire CEO of Meta demonstrated Meta's latest artificial intelligence (AI) model, V-JEPA’s capabilities in the  throwback video where he was playing guitar and singing for his daughter, Maxima.

V-JEPA is a "non-generative model that learns by predicting missing or masked parts of a video in an abstract representation space." It was released on February 15.

OpenAI has also released Sora, its generative videos application.

The AI was able to predict his hand movements while strumming chords, even as it was not trained for that type of analysis.

“Throwback to singing one of Max's favorite songs. I recently tested this video with a new AI model that learns about the world by watching videos. Without being trained to do this, our AI model predicted my hand motion as I strummed chords,” he said.

Apart from predicting missing parts of a real-world video, the AI also predicted concepts from context without fine-tuning.

The predictive analytic algorithm called V-JEPA—short for Video Joint Embedding Predictive Architecture—learns everything from visual media. It is able to anticipate what will happen next in addition to comprehending what is happening in a film. 

V-JEPA takes a self-supervised learning approach and is pre-trained entirely with unlabeled data. Used to adapt the model to a particular task after pre-training, a large portion of the video is masked and only a little part of the context is shown.

 "For example, if the model needs to be able to distinguish between someone putting down a pen, picking up a pen, and pretending to put down a pen but not actually doing it, V-JEPA is quite good compared to previous methods for that high-grade action recognition task," Meta said in a blog post.

The video has garnered 57.6K likes in 3 days.