Our team of five neural networks, OpenAI Five, has started to defeat amateur human teams (including a semi-pro team) at Dota 2: blog.openai.com/openai-five/

A real robot hand, trained with same learning algorithm and code as OpenAI Five, has learned human-like motions to rotate objects: blog.openai.com/learning-dexte…

Reminder: OpenAI Benchmark is today! Watch live at twitch.tv/openai starting 12:30p PT (main event at 1:15p). Follow Greg Brockman and Christy Dennison for livetweeting from the event. Good luck, have fun to both teams! blog.openai.com/openai-five-be…

First OpenAI Five match at The International today! - Streamed on The International's Twitch channels. English: twitch.tv/dota2ti . For newcomers: twitch.tv/dota2ti_newcom… - Held at the end of today's 1p PT matches — likely be around 3p PT. x.com/OpenAI/status/…

OpenAI Five match starting now: twitch.tv/dota2ti ! x.com/OpenAI/status/…

Random Network Distillation: A prediction-based method that achieves state-of-the-art performance on Montezuma’s Revenge -blog.openai.com/reinforcement-…

We've trained an unsupervised language model that can generate coherent paragraphs and perform rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training: blog.openai.com/better-languag…

Lots more work on sparsity in various forms forthcoming...

This is a really cool project I'm happy to have had the chance to help out with.

The fact that primaries are not done via ranked-choice voting is absurd

New video: Simulating an epidemic. What happens when people avoid each other for the most part but still go to a common central location like a store? What if you can track and isolate cases, but 20% slip through the cracks? 50%? And much more. youtu.be/gxAaO2rsdIs

How do we beat COVID-19, *while protecting our mental & financial health?* Epidemiologist @MarcelSalathe & I explain Epidemiology 101, #TestTraceIsolate, #MasksForAll, worst-case scenarios... all with playable simulations! 🕹 Check it out! 😷ncase.me/covid-19/😷 (30 min)

One of the many lessons we should be learning from COVID: the importance of statistics education. Understanding causality and probability is critical to being informed and making wise choices. How much better off would we be if schools replaced trig or calculus with stats?

In case you missed it, here's Aditya Ramesh Justin Johnson and Andrej Karpathy getting into the gritty details of the DALL-E paper: youtu.be/PtdpWC7Sr98 . I hope this kind of interaction becomes the more of the norm in the future.

I have no intention of changing my profile link. help.twitter.com/en/rules-and-p…

The case for 4-bit precision: k-bit Inference Scaling Laws abs: arxiv.org/abs/2212.09720

❤️

OpenAI is nothing without its people

You can find the slides to my NIPS talk on Small World Network Architectures here: supercomputersfordl2017.github.io There were lots of other great talks in that workshop. Thanks to Google (mainly @erich_elsen) for organizing.

How big is the sudden explosion of startups building AI chips? There are 45 of them, and at least five have raised $100 million: nyti.ms/2EEpICX