// Stable Diffusion, Explained //
You've seen the Stable Diffusion AI art all over Twitter.
But how does Stable Diffusion _work_?
A thread explaining diffusion models, latent space representations, and context injection:
1/15
Paper: “Orca: Progressive Learning from Complex Explanation Traces of GPT-4”
Reading and recording here: https://t.co/meh1DyI9LS
Paper here: https://t.co/K2oSVpbNhT
2/2
Community paper reading of “Orca” with @jason_lopatecki of @arizeai is up!
Orca uses “system prompts” to better-finetune small ~10B parameter LLMs from large ones like GPT-4.
This training process extends the small models’ reasoning capabilities.
1/2
ICYMI the recording and transcript of our latest paper reading featuring @brian_a_burns of @ai__pub is up!
Orca leverages rich signals from GPT-4, surpassing state-of-the-art models by over 100% in complex zero-shot reasoning benchmarks.
Check it out:
https://t.co/FmtZWRSVrf
Arize is doing a great paper-reading series every Wedneday focused mostly on modern LLMs.
You can check out past recordings here: https://t.co/a4oIaurjRu
// Paper reading this Wednesday: Orca //
Join me and @jason_lopatecki + the team at @arizeai in reading:
"Orca: Progressive Learning from Complex Explanation Traces of GPT-4"
This Wednesday, 10:15am Pacific!
Discussion is open to all, register here: https://t.co/a4oIaurjRu
Orca is an interesting development in the recent trend of fine-tuning smaller LLMs (~10B parameters) from the output of large ones like GPT-4.
Paper here: https://t.co/K2oSVpbNhT
Workflows to:
- Cluster and debug LLM prompts
- Fine-tune on underfit clusters in latent space
- Observe model drift
0 to almost 1k stars in two weeks!
Homepage: https://t.co/L3WioC7obg
GitHub: https://t.co/YvxncPnE4e
And give it a try with:
pip install arize-phoenix
2/2
// Phoenix: Open-source ML Observability //
@arizeai just released an open-source library for LLMs and deep NNs more generally! It's really cool.
Provides, among other features:
- Visual clustering analysis model interpretability in latent space
...
1/2
Announcing Harvey's Series A led by @sequoia.
Looking forward to working with @gradypb, @charliecurnin and lucky to have the continued support of @OpenAI, @w_conviction, @eladgil and @svangel.
https://t.co/j8Bm0ii6Lh
Harvey just announced a $21M Series A led by Sequoia.
Backed by the OpenAI Startup Fund since August 2022, Harvey is building AI knowledge workers for law.
They're the fastest-growing LLM startup I know of, and are hiring aggressively (below).
https://t.co/UC3vIPD6zB
1/5
To apply to Harvey, either:
1) Fill out this form: https://t.co/wcjE3d1ZeP,
and I will personally review your application + forward to founders.
(I work closely with Harvey as their main outside recruiter.)
4/5
Had a really fun conversation with Louis Bouchard (@Whats_AI) - MILA PhD student, founder of @towards_AI, and host of What's AI (https://t.co/IJiGIEPSSv)!
We spoke about:
- AI media
- AI recruiting & careers
- Breaking into industry or research without a PhD
... and much more!
Are you considering pursuing a PhD or wondering how to get into #machinelearning? Before you take the plunge, listen to this insightful interview with Brian Burns, #PhD candidate at the @UW and founder of the @ai__pub Twitter account.
Some of the best insights 🧵👇