Yang-Hui He

speaker

1,500 appearances 1 recordings 1 series first heard Jan 2025 last heard Jan 2025

Yang-Hui He’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

No recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.

Appearances

newest first · ▶ plays the moment

Theories of Everything with Curt Jaimungal · Yang-Hui He: Why AI Will Permanently Transform How Math Is Done · 3 Jan 2025 podcast

So now you have a a perfectly supervised

perfectly defined binary supervised machine learning problem.

Then you pass it to your your your standard AI, you know.

algorithm.

That they're they're you know just you know out of the box ones.

Nothing uh you you don't even have to tune your particular uh architecture.

Just take your favorite one.

And then do cross validation, you know, the standard stuff, take sample, do the training, you know, and then try to validate this on unseen data.

So if you do this to the model three problem, to to this one.

You'll immediately find that, you know, any any neural network or no whatever base classifiers would do it a hundred percent accuracy, as you should, because see, you'll be really dumb if you didn't, because this is just a line linear transformation.

So even if you have a single neuron that's just doing linear transform, that's good enough to do it.

The prime Q problem, I did some experiment um some oh gosh, I'd be like seven years ago.

And it got 80% accuracy.

And I was like, wow, that's kind of this was a wild moment.

I was like, why is it doing?

I have a I do I don't have a good answer to this.

Um why is it doing 80% accuracy to this?

But how how is it learning?

Maybe it's doing some sieve method, uh, which is kind of interesting.

Showing 701–720 of 1,500 · page 36 of 75