!! @pabbeel and I are building a new AI research lab in SF for Amazon! We’re focused on the remaining major problems to build generally intelligent agents and are looking for a few dozen intrinsically motivated people to join our team and work with the Adept folks here. DM me!

Introducing our new multimodal model Adept FUYU-HEAVY!! Beats Gemini Pro on both MMLU (text-only benchmark) and MMMU (image+text benchmark) Plus cute mascots!

Today we’re opening access to Adept Experiments 🧪, a new way to explore the technology we are developing at Adept. Each experiment is a self-contained mini-tool or demo that showcases a part of our underlying tech. https://t.co/hLghXtMxPx

After much hard work, we’re excited about a series of releases coming up this fall! We’re starting with a small appetizer: we’re open-sourcing Persimmon-8B, the most capable fully permissively-licensed language model with <10B parameters. Details here:

✨ come join me! ✨ We're looking for a product designer to join our team focused on making an AI teammate @AdeptAILabs, on-site in SF. It's a super lovely, amazingly talented, ridiculously humble group, with so much room for impact ❤️

I'm so excited to share what's up next for me I'll be designing a digital AI teammate with the super talented people at @AdeptAILabs! I'm just a week in and couldn't be more excited about the both wildly creative and technically rigorous team. Definitely watch this space 🤩✨

Big round a few days after @basil’s LondonAI talk 😉

We've raised $350M in new funding as part of our Series B led by @generalcatalyst and co-led by @sparkcapital https://t.co/jrQsQnRKOl (1/4)

👋 Join us on 9 March for @londonai! If you're working on AI, this is your jam. If you're interested in AI, you will this jam, ft: 🤖 @vijaybolina of @DeepMind, 👾 @basil of @AdeptAILabs, 🌱 @lorenzphil09 of @Basecamp_Res and mystery speaker :-) https://t.co/c8GMFfz9s9

The AI community expresses sympathy to the crypto community

I'm very excited to have joined @AdeptAILabs! We're building a universal collaborator – it's like an overlay that sits on top of all the software you use. You can hand off tasks to it by just... asking. Below is an early preview of some things it can do!

Finished my final week at @normallystudio, after 7 amazing years. Normally works by combining radical experimentation with… just being kind. This culture leads to impactful and meaningful projects. If you get the chance to work with them, or just meet with them: you should ❤️

I tried my own question using nonsense words – the voodoo incantation worked (without needing the second "and so the answer is..." prompt) @shaneguML @arankomatsuzaki

This is brilliant. The model just needs to be asked nicely to do a good job.

Link to the summary:

Finally: this is a paper showing good results in aligning language models to certain human requirements. Great! But if you log in to your OpenAI account… the model is right there, launched, and is actually the default model. This is v much a product company now (4/꩜) @sama

InstructGPT is 100x smaller (!) than GPT-3 (1.3B params vs 175B) but its outputs are rated higher by human labellers for these tasks. It's somehow nice that the team that kicked off the current GIANT MODELS arms race is also taking steps in a different direction. (3/꩜)

Prompt engineering, one of the most interesting approaches to getting good results from GPT-3, is apparently no longer needed for many tasks with InstructGPT models. Just the question will do. (2/꩜)

A couple of fascinating things in the new alignment paper from @OpenAI (1/꩜)

I would also ✨really✨ like to hear about whatever cool project or hobby you’re in to right now. @ or DM me!