Video Has No Memory. Here's How We Built One. — James Le, TwelveLabs
This talk addresses the challenge of enabling AI models to retain information across extended video content, effectively giving them memory. The presenter discusses the technical hurdles and the architectural solutions developed to overcome these limitations, allowing for more sophisticated video analysis and interaction.
20 min