AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Scaling Laws for Machine Learning
In the last five to 10 years, the very depressing discovery is that this is a mirage. This is not a global optimum. This is a local optimum. It levels off for a bit and then it continues improving. That's amazing for machine learning because that basically precipitated the launch of all these large models. And essentially what it concludes is that there is essentially no limit to how good these models are as long as you can throw enough data at it. So like if you're the Facebook or for the Google or you know, X, Y, or Z Instagram, even like Instagram ads are so Apple, freaking relevant that yeah, Apple for sure, but they're so yeah,