{"id":"oIx44BczzvQ","title":"Illia’s Origin Story: From Google Research to NEAR AI","description":"Before launching NEAR AI, Illia was chasing a bold idea at Google: use natural language to reach the next level of machine intelligence. From cat neurons to transformers to “vibe coding,” here’s how it all started.\n#AI #Founders #MachineLearning #Transformers #TechStory #NearAI","channel_name":"NEAR Protocol","published_at":"2025-07-07T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/oIx44BczzvQ/hqdefault.jpg","transcript_text":"So originally I'm was a researcher. I was always interested in AI and I joined Google research actually because I believed the natural language is the way to get to the next level of intelligence and I saw the paper from Google that was cat neuron where they pre-trained an image model and it learned about cats without any supervision. Super cool. I'm like I want to do that but for text because this is this is how we get to intelligence. So I worked on question answering specifically and so that's kind of how we got to transformers we were like how do we get actually a better model that can scale that can surf you know quickly that can read lots of text uh in a quick way. Now after transformers I wanted to apply that to a product and so I left Google to start nearai and neari was teaching machines to code. Now we call it vibe coding. Back then people called it delusional or scienceiction.","transcript_synced_at":"2026-11-03T04:53:46.350Z","segments":[{"start_seconds":0,"text":"So originally I'm was a researcher. I was always interested in AI and I joined Google research actually because I","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=0s"},{"start_seconds":7,"text":"believed the natural language is the way to get to the next level of intelligence and I saw the paper from Google that was","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=7s"},{"start_seconds":17,"text":"cat neuron where they pre-trained an image model and it learned about cats without any supervision. Super cool. I'm","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=17s"},{"start_seconds":24,"text":"like I want to do that but for text because this is this is how we get to intelligence. So I worked on question answering specifically and so that's","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=24s"},{"start_seconds":32,"text":"kind of how we got to transformers we were like how do we get actually a better model that can scale that can","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=32s"},{"start_seconds":39,"text":"surf you know quickly that can read lots of text uh in a quick way. Now after transformers I wanted to apply that to a","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=39s"},{"start_seconds":46,"text":"product and so I left Google to start nearai and neari was teaching machines to code. Now we call it vibe coding.","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=46s"},{"start_seconds":53,"text":"Back then people called it delusional or scienceiction.","url":"https://www.youtube.com/watch?v=oIx44BczzvQ&t=53s"}],"youtube_url":"https://youtube.com/watch?v=oIx44BczzvQ","url":"https://near-pedia.xyz/videos/oIx44BczzvQ"}