lifelonglearner

Hey, I'm Owen. I think rationality is pretty rad.

lifelonglearner's Comments

April Fools: Announcing LessWrong 3.0 – Now in VR!

Oh no, that was a bad typo. It has now been corrected.

April Fools: Announcing LessWrong 3.0 – Now in VR!

Wow. I'm running 3.1 now, and my laptop's fan isn't running at all. Wild!

April Fools: Announcing LessWrong 3.0 – Now in VR!

What a fantastic product. Reminds me of the 3-d Reddit museum app.

run LessWrong 2.0 by moving everything to a javascript based web-app architecture, so we consider this a natural next step for us to take.

Okay, but actually, though, I'm still hoping for the day where lesswrong.com loads more comparably to greaterwrong.com.

Tessellating Hills: a toy model for demons in imperfect search

Hi, thanks for sharing and experimentally trying out the theory in the previous post! Super cool.

Do you have the code for this up anywhere?

I'm also a little confused by the training procedure. Are you just instantiating a random vector and then doing GD with regards to the loss function you defined? Do the charts show the loss averaged over many random vectors (and function variants)?

Training Regime Day 3: Tips and Tricks

Overall enjoying this series and your take on CFAR-style rationality. Thanks for putting in the time to write this up.

Perceptrons Explained

Michael Nielsen also has some great stuff. Especially his quantum.country and neural networks one.

A Simple Introduction to Neural Networks

Slight ntipick: Simply because logistic isn't actually used in practice anymore, it might be better to start people new to the sequence with a better activation function like reLU or tanh?

When I was first learning this material, due to people mentioning sigmoid a lot, I thought it would be a good default, and then I learned later on that it's actually not the activation function of choice anymore, and hasn't been for a while. (See, for example, Yann LeCun here in 1998 on why normal sigmoid has drawbacks.)

More writeups!

As a university student with ties to EA (and also looking at future opportunities), the EA forum post you linked gave some useful anecdotes to think about. Thank you for sharing the list.

UML IX: Kernels and Boosting

Just wanted to thank you for writing up this series. I've been slowly going through the book on my own. Just finished Chapter 2 and it's awesome to have these notes to review.

The case for lifelogging as life extension

A friend I know actually goes everyday with a GoPro recording his interactions.

Also, I'm wondering if you have thoughts on where to store this preserved information? Making sure that future people have access to it seems like the important part. But obviously just making it all available publicly online for everyone seems too vulnerable. Maybe some sort of dead-man's switch type setup, where it gets made public after you die?

Load More