sarah hong ✨

pursuing the adjacent possible for the next era of human-aligned ai. cofounder @latentspaceai 👩🏻‍💻🇨🇦 @uoftcompsci

[sarah hong ✨ @latentcodes]

We’ll be presenting our paper “Low Distortion Block-Resampling with Spatially Stochastic Networks” today at 9pm PST🧩🏰🎨

Joint work w/ @martinarjovsky @darrylbarnhart @iantbd @LatentSpaceAI!

➡️ paper: https://t.co/ZIUpj4iKNt
➡️ poster: https://t.co/SNbOCs9xqk

#NeurIPS2020

[Garry Tan @garrytan]

Thousands of startups and small businesses are at risk due to the SVB failure. We ask for depositors to be made whole, and for regulation to prevent this catastrophe.

650 founders who employ more than 22,000 people, have already signed.

Please spread the word.

https://t.co/77bYYI4HLT

[christina kim @christinahkim]

THRILLED to share that ChatGPT, our new model optimized for dialog, is now public, free, and accessible to everyone. https://t.co/dyvtHecYbd

[sarah hong ✨ @latentcodes]

@VitalikButerin LLMs are wordcels created through numerous shape rotations

[sarah hong ✨ @latentcodes]

@charles_irl @full_stack_dl congrats charles! 🥞🎉

[sarah hong ✨ @latentcodes]

@tarunchitra @VitalikButerin I heard this before about the state space for Go but… https://t.co/De0eOR9Z5i 😉

But how many rollouts do we actually need if they are high quality? Eg when trying out a new belief, I wonder how many imagined rollouts we actually do to sample potential downstream effects

[sarah hong ✨ @latentcodes]

@tarunchitra @VitalikButerin Do you think your value function would be significantly different if you could do as many rollouts as AlphaGo? 🤔😊

[sarah hong ✨ @latentcodes]

@VitalikButerin I like the Ideological Turing Test for this! I find it pushes me to try to understand the passion underlying a belief system I initially disagree with (more so than steelmanning).

My grand plan is to eventually use this to trick @iantbd into thinking I am a 🤖

[Susan Zhang @suchenzang]

So excited to finally open up access to these models! Couldn't have asked for a better team to do this with: @stephenroller, @NamanGoyal21 (@myleott + @sam_shleifer at the start too)!

[Whyrusleeping @Whyrusleeping]

Human's won't be obsoleted by AGI, we just need better tooling to stay on top.

[sarah hong ✨ @latentcodes]

super excited for this 🎉

watch @DarrylBarnhart's presentation @NVIDIAGTC if you're interested in learning about our approach for scaling LLM with retrieval and appreciate a good library of babel reference 📚🐝🌎

[Latent Space @LatentSpaceAI]

Catch @DarrylBarnhart speak @NVIDIAGTC today about our exciting work @LatentSpaceAI on scaling LLM with retrieval! 📚

There’ll be an open QA in 30 mins: https://t.co/C42yjpjEi1

[Gitcoin @gitcoin]

As we kick off GR13, we’re especially proud to announce:
A 🇺🇦Support for Ukraine🇺🇦 round

We’ve raised $700K to go towards humanitarian aid for Ukrainians

To learn more and get involved, read this 🧵

[sarah hong ✨ @latentcodes]

Me: The greatest love story: Alice & Bob… across time, for all time

Him: Not for all time. It’s until the second law of thermodynamics defeats them — the last lovers — and dissipates them and all their messages into the uniform distribution — the maximum entropy state.

Me: 🙃

[sarah hong ✨ @latentcodes]

@_arohan_ same, same 🍔🍟

<insert recommender system alignment joke>

[sarah hong ✨ @latentcodes]

@PatrickJBlum @pmarca @jordanbpeterson i'm convinced _you_ are the abstract manifestation of chaos 😂

please do newton and darwin next 🙏

https://t.co/X1Udtj3wRF

[Ian Thompson @iantbd]

Wordcel Darwin wrote millions of words in diaries during an expensive journey around the world to get labels for his training data (only to realize it was self-supervision all along…)

Rotator Newton sat at home during a plague rotating apples in his head. (2/4)

[sarah hong ✨ @latentcodes]

@snarkyzk yes! in fact, inspired by the current games, @scottfits is training for the next olympics under my vigilant guidance.

at his exponential rate of progress, he'll be tackling the quad axel next ;) ⛸️⛸️

[Ian Thompson @iantbd]

Here’s another memetic isomorphism:

Crypto is more like physics, and AI is more like biology – not in “biologically inspired” sense, but in the nature of possible solutions.

So using the wordcel / rotator basis, Darwin is a wordcel and Newton is the ultimate rotator. (1/4)

[sarah hong ✨ @latentcodes]

@savvyRL @csmisko Ahh, memories! Remember when you were telling us about your family in Wuhan while also teaching us how to make dumplings? The last gathering shortly before the world turned upside down 🥲

[Sander Dieleman @sedielem]

New blog post: diffusion models are autoencoders!
https://t.co/N0rpUn5CSY

I take a closer look at this connection, inspired by the work of @YSongStanford, @StefanoErmon, @kjgeras, @RandomlyWalking, @gyomalin_ML and @jaschasd, among others!

[sarah hong ✨ @latentcodes]

CPT is much better than the current approach of "hacking" search on top of GPT3.

irrespectively of how it compares to other existing models, the thing I'm worried about is that davinci (175b) does not perform significantly better than ada (300m)...

5/n

[sarah hong ✨ @latentcodes]

a response from CPT author re eval, emphasizing that CPT is one model vs many, fine-tuned specific ones (similar to GPT3 vs task-specific models):

https://t.co/2CLW27GMOG

I do wish there were better ways to evaluate real-world settings! any public datasets / benchmarks? 🤔

4/n

[Arvind Neelakantan @arvind_io]

A thread on how we evaluate our embedding models in OpenAI’s API. We achieve state-of-the-art results in linear probe classification, text search and code search. It’s not fine-tuned, so it works great in the real world — and our customers love it. 1/7