Overclocked Erika @cursecompiler

amphetamine fueled version of @rrika9 Joined February 2010

Tweets

989
Followers

124
Following

33
Likes

5K

Peyman Milanfar @docmilanfar

a year ago

It’s been >20 years since I published my first work on multi-frame super-res (SR) w/ Nhat Nguyen and the late great Gene Golub. Here’s my personal story of SR as I’ve experienced it from theory, to practical algorithms, to deployment in product. In a way it’s been my life’s work

14 110 846 116K 565

Download Image

Erika (rrika9 on 🦋) @rrika9

2 years ago

if there is anything dear to you on this website: - open your browser's devtools - scroll over all the relevant tweets - hit "Save All As HAR" - leave the processing to kind souls later

4 12 86 8K 31

Eric Heitz @eric_heitz

2 years ago

When @_Laurent and I started learning about diffusion models, we were puzzled by the amount of jargon and concepts. So, we derived a model from scratch with our own graphics-people intuitions. Simple derivation, simple implementation, SOTA quality. ggx-research.github.io/publication/20…

30 346 2K 442K 1K

Download Image

Overclocked Erika @cursecompiler

2 years ago

The myth of "consensual" synaptic transmission pre-synaptic axon: I consent post-synaptic dendrite: I consent astrocyte: I don't Isn't there somebody you forgot to ask?

0 0 2 117 1

Download Image

John Nay @johnjnay

2 years ago

Gisting: 26x Compression of LLM Prompts -Trains LLM to compress prompts into smaller sets of "gist" tokens to be reused for compute efficiency -Can be easily trained as part of instruction fine-tuning -FLOPs reductions, time speedups & storage savings arxiv.org/abs/2304.08467

14 158 955 276K 567

Download Image

Aaron Lou @aaron_lou

2 years ago

Presenting Reflected Diffusion Models w/ @StefanoErmon! Diffusion models should reverse a SDE, but common hacks break this. We provide a fix through a general framework. Arxiv: arxiv.org/abs/2304.04740 Github: github.com/louaaron/Refle… Blog: aaronlou.com/blog/2023/refl… 🧵(1/n)

12 80 335 135K 121

Download Gif

Overclocked Erika @cursecompiler

2 years ago

idea: language model tuned for zlib compressible outputs result: "It was a morality core they installed after I flooded the Enrichment Center with a deadly neurotoxin to make me stop flooding the Enrichment Center with a deadly neurotoxin." alternate result: "sorry sorry sorr

0 0 2 163 0

Michael C. Frank @mcxfrank

2 years ago

How do we compare the scale of language learning input for large language models vs. humans? I've been trying to come to grips with recent progress in AI. Let me explain these two illustrations I made to help. 🧵

50 362 2K 2.6M 958

Download Image

Overclocked Erika @cursecompiler

2 years ago

I kind of want to see a transformer trained on machine code, mostly to measure information content. Like, does this 20mb exe really need to be this big? kind of like keeping an eye on the size after gzip, but more advanced

0 0 0 54 0

Overclocked Erika @cursecompiler

2 years ago

remember when microsoft added all the powerpoint slide transitions to IE6, so their html export would work?

Maxi Ferreira @charca

2 years ago

remember when microsoft added all the powerpoint slide transitions to IE6, so their html export would work?

34 328 3K 351K 1K

Download Video

0 0 0 68 0

Overclocked Erika @cursecompiler

2 years ago

downloaded 19GiB of TF and dependencies just to get > Ignoring visible gpu device with AMDGPU version : gfx1035. The supported AMDGPU versions are gfx1030, gfx900, gfx906, gfx908, gfx90a.

0 0 0 62 0

Guillaume Dumas @introspection

4 years ago

Joint Attention for Multi-Agent Coordination and Social Learning arxiv.org/abs/2104.07750… v/@SamBolotta #SocialNeuroAI

1 17 43 0 7

Download Image

hardmaru @hardmaru

6 years ago

Effective, interpretable algorithms for curiosity automatically discovered by evolutionary search I like this direction. Things like imitation, curiosity, memory, world models, learning etc., are emergent properties of evolution, optimizing for survival. arxiv.org/abs/2003.05325

2 101 398 0 99

Download Image

Jakob Voigts @jvoigts

4 years ago

How do neurons actually sum up inputs? Turns out, it depends a lot on where the synapses are! Team effort by Lafourcade et al. in Harnett lab finds parts of dendrites are linear➕ others nonlinear❌, depending on location on dendrites, via AMPA/NMDA ratio. authors.elsevier.com/a/1eb%7EW3BtfH…