Why AI Actually Works (Without the Hype)

257 views· 6 likes· 4:01· Jul 3, 2025

🛍️ Products Mentioned (4)

Timestamps: 0:00 Intro 0:27 First Analogy to understand 1:26 How LLMs work 1:46 How it all started - Attention is all you need paper 2:18 Understanding ≠ Building & Second Analogy 3:02 Interpretability 3:16 Golden gate research of Claude 3:49 See you again ------------------------------------------------------------------------------------------------------------------------- Summary: 🔍 How AI Actually Works — Transformers, LLMs, and the Mystery of the Golden Gate Bridge In this video, we break down how modern AI models like ChatGPT, Claude, and Gemini really work — and why they’re not just guessing, but storing real-world concepts inside their digital “brains.” We trace the journey from the “Attention Is All You Need” paper that introduced the Transformer architecture, to how Large Language Models (LLMs) predict the next word using probabilities. But here’s the twist: researchers still don’t fully understand why these models work so well. That’s where AI interpretability comes in — the science of looking inside these black boxes to see what's really going on. One mind-blowing example? Anthropic found a specific feature inside their model that represented the Golden Gate Bridge. When they amplified it, the model wouldn’t shut up about it — proving that LLMs aren’t just random math… they store actual concepts. 🔥 Topics covered: - What are Large Language Models (LLMs)? - How Transformers changed the AI game - The origins of modern AI from machine translation - What is AI interpretability and why it matters - Anthropic’s discovery of real-world concepts in AI models If you're curious about how AI works under the hood — or how researchers are trying to make sense of it — this is the video for you. #ai #aiagents #artificialintelligence -------------------------------------------------------------------------------------------------------------------------- Resources mentioned: - Attention is All You Need: https://arxiv.org/pdf/1706.03762 - Emergent abilities of LLMs: https://arxiv.org/pdf/2206.07682 - Golden Gate reference: https://www.anthropic.com/news/golden-gate-claude ---------------------------------------------------------------------------------------------------------------------- This video is made for AI and tech enthusiast. Hi, in this channel I generally talk about AI, tech, software and hardware. I am trying to make topics like AI, AI agents etc simpler for people. I talk about technology, new tools, learning resources and explain different topics. Catch me on LinkedIn: https://www.linkedin.com/in/heysourin/ GitHub: https://github.com/heysourin

🎬 More from NextHertz AI With Sourin