{"id":"j2PzsKbUtn4","title":"🧠 AI Mind-Bender: Infinite Learning in Finite Models? 🤖","description":"Alex and Illia discuss NEAR AI brainstorming sessions.\n\nFunction G: Model + Data Batch = Enhanced Model\nCan models infinitely expand knowledge?\nThe twist: Size constraints weren't specified!\n\n#AILearning #MachineLearning #TechPhilosophy #DataScience #NEARai #NEARuserownedAI\n\n🔔 Subscribe for more mind-expanding tech discussions!\n\nJoin NEAR's community:\nWebsite: https://near.org/\nNEAR AI: https://near.ai/\nTwitter:   / nearprotocol  \nBlog: https://near.org/blog/\nMedium:   / nearprotocol  \nGitHub: https://github.com/near \nDev Docs: https://docs.near.org/\n#NEAR #nearprotocol #NEARAI","channel_name":"NEAR Protocol","published_at":"2024-07-03T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/j2PzsKbUtn4/hqdefault.jpg","transcript_text":"Yesterday, Illia out of nowhere stood up and philosophically said, I want the Function G. Such that if you feed into it a model and a batch of data using a new model such that the new model knows everything, that the original model knew plus information from the batch. Right. And feeling very smart. I was like, well, this is obviously impossible because then by doing it repeatedly, you can learn infinite amount of data until the final size of the model. But I guess obviously there is no nothing in this problem statement towards requiring the model to remain of the same size. Right. And this is sort of that's a version of the Function G. feeling smart, being smart is not quite the same thing. But the point of this is then you know, you're able indeed to maintain the knowledge and skills but also potentially grow them.","transcript_synced_at":"2026-11-03T05:46:22.295Z","segments":[{"start_seconds":0,"text":"Yesterday, Illia out of nowhere stood up and philosophically said, I want the Function G.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=0s"},{"start_seconds":9,"text":"Such that if you feed into it a model and a batch of data","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=9s"},{"start_seconds":16,"text":"using a new model such that the new model knows everything, that the original model knew plus information from the batch. Right. And feeling very smart.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=16s"},{"start_seconds":26,"text":"I was like, well, this is obviously impossible because then by doing it repeatedly, you can learn infinite amount of data until the final size of the model.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=26s"},{"start_seconds":35,"text":"But I guess obviously there is no nothing in this problem statement towards requiring the model to remain of the same size. Right.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=35s"},{"start_seconds":41,"text":"And this is sort of that's a version of the Function G. feeling smart, being smart is not quite the same thing.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=41s"},{"start_seconds":49,"text":"But the point of this is then you know, you're able indeed to maintain the knowledge and skills but also potentially grow them.","url":"https://www.youtube.com/watch?v=j2PzsKbUtn4&t=49s"}],"youtube_url":"https://youtube.com/watch?v=j2PzsKbUtn4","url":"https://near-pedia.xyz/videos/j2PzsKbUtn4"}