jackson1372

I don't have sources, it's just a rumor.

It's an open secret that Mistral finetunes on GPT outputs

claude.ai

The article doesn't mention two great features on iOS: - Move cursor by tap-and-hold on soace bar - tap-tap for word selection and tap-tap-tap for sentence selection.

I use these two constantly

Claude Instant 1.1 is a better comparison for the price/performance of GPT-3.5

GRE writing: 5 for Claude vs 4 for GPT-4

Bar exam: 76.5 for Claude vs 75.7 for GPT-4

Isn't the explanation for this that the world actually does work in some way or other and that it's not just infinite chaos and so if you keep throwing parameters at some problem, you will eventually stumble upon the "real" structure, but that's no guarantee of when that occurs, and with which parameters?

The reason you want to over-parameterize your model is that it protects you from "bad bounce" learning trajectories. You effectively spread out your overfitting risk until it's pretty close to 0.

Or at least that's the way I like to think of it.

The next step is to better compress the resulting model in a simpler, less computationally costly network.

If you read between the lines, it's clear that Spotify was able to make a normal watch app. But they wanted to make one that had special functionality not yet allowed by Apple, for any app, not just Spotify.