some news: i quit MSFT last week after 3.5 years with the inflection/microsoft AI crew -- lots of memorable times. i'm taking some time to reconnect w/ old colleagues and friends and its been great. if you are reading this and want to catch up, DM me!!

Excited to announce that we’ve raised $1.3B to build one of the largest clusters in the world and turbocharge the creation of Pi, your personal AI. https://t.co/p5AfRXGPan

It’s a big week! We’ve raised $1.3 billion and are building the world’s largest AI cluster (22k H100s). We’re grateful for our investors and new funding that will help us accelerate our mission to make personal AI available to every person in the world.

We have amazing results to announce! Inflection-1 is our new best-in-class LLM powering Pi, outperforming GPT-3.5, Llama and PALM-540B on major benchmarks commonly used for comparing LLMs.

Kullback-Leibler divergence is not the same as Leibler-Kullback divergence

I've ported @rewonfc's very deep VAE ( https://t.co/0lQp9BQgcI) from PyTorch to JAX/Flax! Hope other JAX users find this SOTA VAE useful as a forkable baseline... https://t.co/rJno4lcLSO.

My post on SotA image generative models was released 🥳 Featured 7 notable recent papers with emphasis on: - VD-VAE - VAE + discriminator (e.g. VQGAN, DC-VAE) - Diffusion models (e.g. DDPMv2) Plus some notes on scaling (e.g. DALL-E) and evaluation. https://t.co/uOV6jVwdmA

IMO, best empirical proof to date that AI can be creative. After this sinks in, will there be any naysayers left?

Congrats to Aditya and the rest of the team for an awesome release!

It is easy to write a program but it is difficult to create a machine that will read those lines. (Was looking through my journal, found this gpt-3 generation conditioned on haikus)

Really thought-provoking work -- congrats to the authors @poolio @YSongStanford @dpkingma and more!

It breaks my 💚 when researchers tell me that VAEs don't work. My first typical question is "did you try hierarchial VAE or vanilla VAE?", the answer is usually vanilla VAE. VAEs work much better with hierarchical structures. NVAEs and this work take this to the extreme!

Posted my first paper on arXiv💥🙌 GPT-f is a Transformer-based automated theorem prover. We show that Transformer + Search is suitable to formal reasoning and continuous self-improvement 🦾 https://t.co/VllDcCV3Kc

Our base model is a Sparse Transformer. If we make it bigger and train for a while with this augmentation, it results in both very high likelihoods (2.55-2.65 bpd on CIFAR-10) and also samples equal/better than most GANs (as measured by FID). Code here:

Thanks if you came to our ICML poster on Distribution Augmentation. The zoom discussion was way more fun/interesting than I expected! TLDR of our work: use powerful data aug in your generative model by conditioning it on the aug. Improves samples + likelihoods considerably.

I keep seeing all kinds of crazy reports about people's experiences with GPT-3, so I figured that I'd collect a thread of them.

Excited to share what I've been working on with @AlecRad, @rewonfc, @ilyasut and others!

Hot Tub Christmas (with GPT-2 lyrics selected by @rewonfc ) is my own favorite -- even though the model can't figure out what's going on in the completely terrible intro. Our weird but fun new holiday tradition!

One of my favorites from @OpenAI's jukebox: 'Lose Yourself' re-rendered by Kanye https://t.co/PlYKbaEjon