How Does AI Actually Know Anything

#NeuralNetworks #LLM #HowAIWorks
How does AI work? How do ChatGPT and other large language models (LLMs) turn your question into a useful answer? Go inside the neural network to see how words become numbers, attention connects them, and training builds the patterns that support knowledge and reasoning.

This documentary follows the models behind ChatGPT, Claude and Gemini, from tokens and learned weights (and why they are called weights) to planning ahead, reasoning steps and the limits of our understanding. A simple question about boiling an egg is our starting example; the same journey leads to much harder tasks in mathematics and coding.

PART ONE, the hardware journey: What Happens Inside an AI Chip When You Ask a Question • What Happens Inside an AI Chip When You As...
Playlist, How AI Works, From Chip to Answer: • How AI Works, From Chip to Answer

CHAPTERS
0:00 How does an AI know how long to boil an egg?
0:52 One word at a time
2:33 Where our first film stopped
3:45 Why "weights"? Brain cells and neural networks
4:48 It can't read: words become addresses
6:45 The words look back (attention)
9:05 The floor: one number
10:29 Who set them? Nobody did
12:15 Why guessing is enough
14:53 They looked inside
17:41 Paper and pencil: what "thinking" is
19:17 Practice, and a gold medal
21:57 It can't tell you how it adds
23:19 The answer