zlacker

Every time Karpathy quits his job, the field takes a leap forward because he makes some fantastic educational resource in his free time.

>>Imnimo+(OP)
Examples? (I'm not that familiar with field.)

>>skybri+v4
The most recent is this, which I believe was made after he left Tesla:

https://github.com/karpathy/nanoGPT

And it's accompanying video series:

https://karpathy.ai/zero-to-hero.html

Another example (although I honestly don't remember if he made this one between jobs) is: https://github.com/karpathy/micrograd

>>skybri+v4
Neural Networks: from zero to hero

https://karpathy.ai/zero-to-hero.html

>>Imnimo+(OP)
yup, I hope we get awsome open source-related content now.

>>skybri+v4
Andrej Karpathy is badmephisto, a name you might have heard of if you're into cubing.

http://badmephisto.com/

>>weinzi+4e
You just blew my mind, I used to hang around this site a lot ~10 years ago and never would have made the connection

>>weinzi+4e
i saw this comment and literally shouted “holy fucking shit” — i use zz method, but i came across lots of his resources before!!!

>>weinzi+4e
Wow, that's a connection between the eras of my life I would not have thought existed. Thank you.

>>weinzi+4e
This blew my mind as well!! I never thought one of my favorite programmers would share a similar hobby haha

>>weinzi+4e
Oh wow this blew my mind

>>weinzi+4e
I haven't been that surprised by something in a long time. Wow that is crazy. I made a little unfinished 3d Rubik's Cube site for fun a while back and the about section includes a link to his channel and some other older cubing channels. https://rubie-cubie.vercel.app/

>>weinzi+4e
Holy crap!! I never knew that. I watched this guy so much in 5th grade, he helped me get my 3x3 time down to 8 seconds.

This is insane

>>skybri+v4
My master's was in Convolutional NNs for language processing. I had zero prior knowledge and my advisor recommended I watch Karpathy's lectures[1] to get up to speed

[1] https://youtube.com/playlist?list=PLkt2uSq6rBVctENoVBg1TpCC7...

>>weinzi+4e
I learned F2L from him in 2009!

>>skybri+v4
And he was teaching CS231n in Stanford in 2016

https://www.youtube.com/watch?v=NfnWJUyUJYU&list=PLkt2uSq6rB...

>>weinzi+4e
10 years ago: https://youtu.be/WhPjlnWbtS8?feature=shared&t=359

>>Copenj+wT
Wow - that earnestly gave me goosebumps. I'm a Googler myself and it's humbling seeing him casually describe, 10 years ago, a technology the industry was still in the early stages of developing, which has since taken the world by storm. What a rockstar.

>>joenot+401
Neural networks were already big 10 years ago, you have to go back 15 years to see before they started being popular.

From wikipedia:

> Between 2009 and 2012, ANNs began winning prizes in image recognition contests, approaching human level performance on various tasks, initially in pattern recognition and handwriting recognition.

That was when Neural networks became a big thing every tech person knew about, 2014 it was already in full swing and you had neural networks do stuff everywhere, like recognizing faces or classifying images.

>>Jensso+Cw1
NN were already a casual topic in my high school computer science class more than 20 years ago. I've always assumed they were already fairly common by that point. (~2000)

>>altint+NA1
They were a scientific curiosity at that point, the widespread use in the industry happened around 10-15 years ago.

>>Copenj+wT
Him solving a rubix cube and riding a bike at the same time is a pretty impressive demonstration of motor coordination

>>Jensso+Cw1
For reference, AlphaGo defeated Lee Sedol in March 2016, which was another reminder to a broader non-tech audience how far things had gotten.

>>weinzi+4e
WHAT!

>>altint+NA1
They were and they were in use, for instance in character recognition. They just hadn't had their breakout success yet.

>>altint+NA1
They were known in the field but had a reputation for being too slow. I remember a couple of early 2000s NIPS (now NeurIPS) people commenting about what a shame it was that NN were computationally infeasible, which was true in the era before GPUs took off.

>>kbelde+6o2
Neural networks weren't the best models for character recognition, their breakout success was when they started being the best at recognize characters and other images which happened in the late 00's. OCR before then was really bad.

Might be hard to imagine today but back then OCR and image recognition was typically done with normal statistical regression models, and the neural networks they had then were worse than those.

>>Copenj+wT
Thank you for sharing this link. The Gates building has changed so much over the last decade, it was nice to see it as it once was.

>>joenot+401
Neural networks were mentioned -- not particularly often, but now and then -- in such non-rarefied publications as BYTE Magazine in the 1980s and 90s, AFAICR.