Christopher Wolff

Safety research @ OpenAI. Previously image generation @ Google DeepMind.

[Micah Carroll @MicahCarroll]

We are hiring!

This may be the best time ever to join: we are incredibly bandwidth bottlenecked and there is a lot of support for almost any impactful misalignment work you can think of. I think we also have a pretty good epistemic environment and a very fun team

Topics: all-things monitoring, misalignment & monitorability assessments, misalignment science, helping set up 3P auditing (e.g. recent Redwood collab), communicating risk externally (system cards/blogs, safety cases, etc)

RSI/misalignment subteam: https://t.co/YHLMCYwWOe

[OpenAI @OpenAI]

Introducing GPT-Red

An internal automated red teamer on a mission to find our models’ prompt injection vulnerabilities at scale, helping us build stronger defenses before wider deployment.

https://t.co/GxnmxxcpSk

[Christopher Wolff @cwolffff]

@NoamShazeer Welcome!

[Google DeepMind @GoogleDeepMind]

Image generation with Gemini just got a bananas upgrade and is the new state-of-the-art image generation and editing model. 🤯

From photorealistic masterpieces to mind-bending fantasy worlds, you can now natively produce, edit and refine visuals with new levels of reasoning, control and creativity.

A quick dive into Gemini 2.5 Flash’s capabilities 🧵

[Demis Hassabis @demishassabis]

The Gemini era is here. Thrilled to launch Gemini 1.0, our most capable & general AI model. Built to be natively multimodal, it can understand many types of info. Efficient & flexible, it comes in 3 sizes each best-in-class & optimized for different uses https://t.co/VUu1277bC2

[Christopher Wolff @cwolffff]

@kramnikstudent @GMJacobAagaard @askavrukh which opening line is this from? iirc i had this exact position before but can't remember

[Christopher Wolff @cwolffff]

@bneyshabur @DeepMind @OriolVinyalsML Welcome! 🙃

[Andrew Fitzgibbon @Awfidius]

So many transformer explanations want me to intuitively understand what's happening. I just want to understand what the code does. This is 25 lines of pure functional code which helped me, and might others.

https://t.co/0LstU86SpO

[Christopher Wolff @cwolffff]

@tanmingxing How many values of epsilon do you typically need to try out to see improvements over CE? And how do the sub-optimal values of epsilon compare?

[Christopher Wolff @cwolffff]

@sahiralsaid @stanfordnlp ty!!

[Christopher Wolff @cwolffff]

@Graftekkel I'm ~2400 but haven't done much coaching before. Feel free to shoot me a pm if you wanna talk more!

[Lee Sharpe @LeeSharpeNFL]

A friend just sent me this ten-year-old prize structure for a Starcraft tournament, and I can't stop laughing.

[Christopher Wolff @cwolffff]

@lexfridman Technically, you could add an extra player that absorbs the positive utility of everyone when they die

[Christopher Wolff @cwolffff]

@polynoamial @PrimordialAA @caleb_banister @ryanzarick Do you think that having a commercial implementation of Pluribus would be helpful to poker players? To my understanding, this would greatly increase computation time, avoid the need for preflop ranges, and enable analysis of multiway spots.

[Christopher Wolff @cwolffff]

@OpenAI Will you have an internship program in 2021?

[Christopher Wolff @cwolffff]

@poker_now_club The updated speed is much better! I think I'd still prefer a near instant transition, but this is great.

[Christopher Wolff @cwolffff]

@poker_now_club I'm a big fan of the site, but one of the main reasons my friends and I have our homegame elsewhere is that there's a huge delay after someone takes an action. It takes forever to fold around the table, whereas on most other sites it's instantaneous.

[Christopher Wolff @cwolffff]

@lipoker_io Some other suggestions:
- An optional decision timer would be awesome.
- Preflop raise sizes in bb don't match the % amounts. A 100% raise should be 3bb heads up, not 1.5bb.
- The ability to choose your seat when sitting down (currently just picks the next open one)