Dr. Fei Fei Li
Your Kid Needs 15 Million Reps
Inspired by Dr. Fei Fei Li episode
Dr. Fei-Fei Li built ImageNet with 15 million images because AI needed massive data to learn. Your kid? Same. You showed them how to use a spoon twice and wonder why they're still shoveling yogurt with their fist.
Fei-Fei realized the problem with early AI wasn't the algorithms—it was the data. Neural networks needed to see cats 10,000 different ways before they could recognize "cat." Humans learn the exact same way. A toddler doesn't master stairs after three attempts. They need hundreds.
Your kid has now asked "why" about the same bedtime rule 47 times this week. They're not torturing you for fun. They're running experiments. Gathering data. Building their neural network.
You taught them how to put on shoes on Monday. By Friday they're crying because the Velcro is "wrong" and they can't remember which foot is which. You didn't fail as a teacher. You just forgot they need the ImageNet-scale repetition. Hundreds of times. Thousands, even.
The AI that changed the world needed 15 million training examples. Your tiny human is working with similar constraints, just with a way better power efficiency rating.
