Emad
Only two countries globally currently have orgs with the capability to train GPT-4+ models.

The USA and... the UK

🤔
tweet

Aran Komatsuzaki
Self-Refine: Iterative Refinement with Self-Feedback

Presents a novel approach that allows LLMs to iteratively refine outputs and incorporate feedback along multiple dimensions to improve performance on diverse tasks.

proj: https://t.co/GQHZ7keE76
abs: https://t.co/LLkvsQG5Ic https://t.co/htXazX451l
tweet

Yann LeCun
Some folks say "I'm scared of AGI"

Are they scared of flying?
No!
Not because airplanes can't crash.
But because engineers have made airliners very safe.

Why would AI be any different?
Why should AI engineers be more scared of AI than aircraft engineers were scared of flying?
tweet

Yohei
The thread broke, but here's a simple example of how "relevant context" (most relevant tasks from past) can be provided to current task.

Allowing the autonomous agent to continue generating novel ideas and next steps - that wouldn't fit within a context window. https://t.co/v5lC9DxFSV
By task 23 here, you see that the relevant context becomes increasingly relevant.

You can see how this can help this next task produce new unique content on top of this. https://t.co/mlYWbxq1UB - Yohei
tweet

Tom Morgan
Many actual LOLs https://t.co/6wMsL8Y3YH
Probably the best thing you'll see today.

In 2017, a group of developers hilariously competed for who could create worst volume control interface in the world.

The results 🧵

1/22 - 0xDesigner
tweet

Stephanie 🏹
i am either overwhelmed or bored there is no in between
tweet

Aran Komatsuzaki
Where are we in the search for an Artificial Visual Cortex for Embodied Intelligence?

Presents the largest empirical study of pre-trained visual representations or visual foundation models for Embodied AI

proj: https://t.co/kw3rtsGSE2
abs: https://t.co/9cZHb7rKXA https://t.co/fLdjaigKup
tweet

Stephen Young
RT @ylecun: People should pay considerably more attention to this simple fact. https://t.co/LNCHFIqe4O
@ylecun It’s weird how the people who who are terrified of AGI are rarely the people who actually build AI models - Jay Kess
tweet

Aran Komatsuzaki
∞-Diff: Infinite Resolution Diffusion with Subsampled Mollified States

Achieves significantly higher sample quality as well as being able to effectively scale to higher resolutions than the training data while retaining detail.

https://t.co/EgPwv3QtgR https://t.co/uKR5BHwazn
tweet

Stephanie 🏹
RT @PopBase: LE SSERAFIM announces their debut studio album ‘UNFORGIVEN.’ Out Monday, May 1st. https://t.co/uc2Rnc2QMt
tweet

Yohei
Added a new "Relevant Context" section that shows up in the "Task Execution" stage.

This is to show how we're using @Pinecone vector search to provide memory from past tasks into the current task being executed.

It's empty on step 1. https://t.co/bvm85D9Qoi
tweet

Yann LeCun
RT @ReligionProf: Lots of people don't understand what ChatGPT is and what it's designed to do. They're surprised it makes things up and says things that are not factual. This article may help.

https://t.co/bCTgMtiUgx
tweet