Ben Gilbert

building high-level features using large-scale, unsupervised learning.

But everyone just calls it the cat paper.

You talk to anyone at Google, you talk to anyone at AI, they're like, oh yeah, the cat paper.

What they did was they trained a large nine-layer neural network to recognize cats from unlabeled frames of YouTube videos using 16,000 CPU cores on 1,000 different machines.

And listeners, just to underscore how seminal this is, we actually talked with Sundar in prep for the episode, and he cited seeing the cat paper come across his desk as one of the key moments that sticks in his brain in Google's story.

Yeah.

It proved that large neural networks could actually learn meaningful patterns without supervision and without labeled data.

And not only that, it could run on a distributed system that Google built to

actually make it work on their infrastructure.

And that is a huge unlock of the whole thing.

Google's got this big infrastructure asset.

Can we take this theoretical computer science idea that the researchers have come up with and use dist belief to actually run it on our system?

Pattern recognizers and data.

Right.

And whether you're searching for a video or they're trying to figure out what video to recommend next, they need to know what the video is about.

If you can answer the question cat or not a cat, you can answer a whole lot more questions too.

It's so crazy.

I mean, this is the craziest thing about unlabeled data, unsupervised learning, that a system can learn what a cat is without ever being explicitly told what a cat is.

And that there's a cat neuron.

Not to mention porn filtering, explicit content filtering.