I think so it's like a new pipeline for them where they compress the videos and store them in a separate ANN that acts as the storage device instead for any other ANN, so they can modify the videos faster/ store them and use them without decompressing them, and I "think" while ANNs are fed already shrunken data to train on and then further compress it into learned features in the neural network, Deepmind goes further to have it fed the data already compressed also I am guessing, and so it is fed data that looks maybe just a bitttttt lower quality than otherwise fed. It sounds to me like they are training easier then, like 700x easier/ faster, by compressing lossyly the videos but maintaining the quality. I mean lossly compressed as a code would be small due to lossly and due to being a code - a code reversed to decompress into the video.
Like, if they can now train on hour long videos, then the 5 second videos they did must be easier to do. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tb7e779c9b08a054e-Ma9ef86bfce469c8663f4ac83 Delivery options: https://agi.topicbox.com/groups/agi/subscription
