Will be joining Microsoft AI to build the next generation of consumer AI!!! 🚀🚀

Today at Inflection we are announcing some important updates. A new phase for the company begins now.
Read more here:
inflection.ai/the-new-inflec…

Really proud of the team 🚀🚀

Excited about the future of Inflection AI !

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. inflection.ai/inflection-1

We are hiring! We develop our own LLMs entirely in house. Our models are currently at SOTA across a very wide range of tasks.

If you want to work on some of the best and largest in-production AI systems in the world, just get in touch...

inflection.ai/careers

Happy to share Pi with the world!

Grateful to be part of such an amazing team!

Minerva author on AI solving math:
- IMO gold by 2026 seems reasonable
- superhuman math in 2026 not crazy
- auto-formalizing is unimpressive to mathematicians as most important theorems are hard to formalize

Happy to release our work on Language Model Cascades. Read on to learn how we can unify existing methods for interacting models (scratchpad/chain of thought, verifiers, tool-use, …) in the language of probabilistic programming.

paper: arxiv.org/abs/2207.10342

@ ICML workshops til Sunday!

Come by beyond-bayes.github.io workshop Friday @ 9:40am for our talk, with posters @ 5pm.

You'll learn how probabilistic programming lets us formalize models talking to models ('model cascades'), unifying many approaches to prompting and inference.

After almost three great years at Google, I decided to move on to my next adventure at Inflection AI to work on conversational AI. Thanks to everyone at Blueshift and Brain for such great times!

✨🧮 Am getting a kick out of reviewing the examples in Minerva's sample explorer:

minerva-demo.github.io/#category=Phys…

Anybody want to take bets on how long it will take to have an automated Physics, Chemistry, or Calculus homework checker as a service? 😂

📄: ai.googleblog.com/2022/06/minerv…

AI lawyer (student?) coming soon?

Interestingly, forecasters' biggest miss was on the MATH dataset, where alewkowycz Ethan Dyer and others set a record of 50.3% on the very last day of June! One day made a huge difference.

With current evaluations, the closest application seems education. Really curious what these models can do in terms of knowledge generation. Even if these models are interpolating in a large space, there are many holes to fill in the scientific knowledge graph!

The paper focuses on quantifiable problem solving, but 🦉does great at explaining technical concepts. It has read all of the arXiv after all.

Curious about REINFORCE? Just prompt it to write a paper on it:
`\section{A derivation of the score function gradient estimator}`

1/ Super excited to introduce #Minerva 🦉(goo.gle/3yGpTN7). Minerva was trained on math and science found on the web and can solve many multi-step quantitative reasoning problems.

My buddy alewkowycz's ultimate revenge. Leave physics and then work on putting me out of a job as well!

Large language models continuing their bit surprisingly rapid advances, here in solving math/STEM problems, without substantial architecture modifications or paradigm shifts. 'The main novelty of this paper is a large training dataset', and fine-tuning on top of PaLM 540B.

Super excited and proud of my colleagues that today announced #Minerva 🦉(goo.gle/3yGpTN7). Minerva was trained on math and science found on the web and can solve many multi-step quantitative reasoning problems.