Andrew Huberman· PhD
We have a thousand trillion running that computation. And that computation is much more sophisticated than the computation that's happening when you train a large language network. Most of your listeners will have heard of artificial intelligence, machine learning, have some basic idea, but we know how that works because we built it. And it doesn't work how you and I work. It's a big, giant global learning signal and we just keep running around feeding it back in, and everybody learns, this is the right answer, and it threads through the entire network. Right and wrong is told to every single neuron in the network. It's something called gradient descent. And that process of improving the network, making it better and better and better at predicting has led to the revolution we've seen in artificial intelligence. What's cool is that our network is cooler. It's way bigger. ChatGPT has 540 billion weights. We are 150 trillion. So, that makes each of us 500 times bigger in scale.