EncodeThis Streams

AI research guy ( who codes and streams at Scientist | Neuro Ph.D | (neo)vim user.

[EncodeThis Streams @EncodeThat]

That's all, folks. Will leave ya'll with some nice feedback I got today: "Thanks for streaming! Would definitely recommend this. This is the real deal and it will give a good impression of how difficult ML really is, even for simple problems being solved by an expert!"

[EncodeThis Streams @EncodeThat]

It's been a while. Streaming at https://t.co/eKeBZcPwNG - trying to solve "Escape From The Aliens In Outer Space" using PyMC

[EncodeThis Streams @EncodeThat]

It's been a while awhile, going to do some Advent of Code 2023 with machine learning / neural networks in jax. Come hang out, there's a 75% chance this doesn't work https://t.co/oM5brllb0M 😂 #adventofcode #jax #jaxline #haiku #weightsandbiases #deeplearning

[EncodeThis Streams @EncodeThat]

It's been a while, going to do some Advent of Code 2022 with machine learning / graph neural networks in jax. Come hang out, there's a 75% chance this doesn't work https://t.co/oM5brllb0M 😂 #adventofcode #jax #jaxline #haiku #weightsandbiases #deeplearning #graphnets #jraph

[EncodeThis Streams @EncodeThat]

I'll be going live in about 20 minutes, come check out some Advent of Code 2021 + deep learning. Probably try out some transformers. #adventofcode #jax #jaxline #haiku #weightsandbiases #python #deeplearning #dnn

https://t.co/eKeBZcQ4De

[EncodeThis Streams @EncodeThat]

Been a minute, gunna do some neovim 0.8 setup + getting started on Advent of Code 2021 with machine learning.

https://t.co/eKeBZcPwNG

[Stephanie Chan @scychan_brains]

New paper 🥳: Transformer inductive biases!

Transformers generalize differently from information stored in:
‣ weights - mostly "rule-based"
‣ context - mostly "exemplar-based"

This effect depends on (a) the training data (b) the size of the transformer
🧵⬇️

[EncodeThis Streams @EncodeThat]

@TheIdOfAlan I do this all the time. When I finally realize what the issue is, I like to move my cursor all the way to one side and then take a running start to try to cross the gap. Hasn't worked yet, but one day...

[EncodeThis Streams @EncodeThat]

It's official, these are my new favorite #neovim mappings:

<SPACE>j vim.lsp.diagnostic.goto_next()
<SPACE>k vim.lsp.diagnostic.goto_prev()

[EncodeThis Streams @EncodeThat]

@Malkyne Every time I think about moving back to the US, I think about the utter scam that is American health insurance and stay right where I am.

[EncodeThis Streams @EncodeThat]

@ThePrimeagen Honestly I'm pretty sure this is the closest I've come to understanding what monads are.

[EncodeThis Streams @EncodeThat]

Looks like solid work, but did they really just call it "DeepNet"? So generic. I suspect it's not very futureproof either.

[Sepp Hochreiter @HochreiterSepp]

ArXiv https://t.co/V3HhzAZgmA: Transformers up to 1,000 layers using a new normalization technique. Outperforms baselines on machine translation IWSLT-14 (De-En) and WMT-17 (En-De). New SOTA on multilingual benchmark with 7,482 translation directions with a 200-layer model.

[EncodeThis Streams @EncodeThat]

@zacharylipton I had a related beef regarding this very example just yesterday: someone tweeted a paper that used "ERM" abbrev. in the title and abstract without ever saying what ERM stands for in the abstract! 👎

[EncodeThis Streams @EncodeThat]

@DThompsonDev @mvortiz "A foolish consistency is the hobgoblin of little minds." - Ralph Waldo Emerson

I used to be all about this quote when I was younger.

[EncodeThis Streams @EncodeThat]

@PhilCulliton @rasbt They should call them enlightenment matrices!

[EncodeThis Streams @EncodeThat]

@ThePrimeagen "More is more" is the phrase I like for the disproportionate benefit you get from being exposed to a variety of perspectives.

[EncodeThis Streams @EncodeThat]

I think my new work phone is trying to tell me something 😅

[EncodeThis Streams @EncodeThat]

@mvortiz This immediately put "Oye cabron" in my head. No clue why, not even a particularly good insult lol

[EncodeThis Streams @EncodeThat]

What's a concept in #deeplearning that you're struggling to understand?

Let's make a list and help each other through these topics.

[EncodeThis Streams @EncodeThat]

@MelkeyDev @ched_dev Speaking from experience, that's a double-edged sword 😅

[EncodeThis Streams @EncodeThat]

@RodOrtJose Just finished doing this on stream literally 10 minutes ago for the Neural Arithmetic Logic Units paper...If ya'll are interested:
https://t.co/izrO712uWv