Founder of @FriendsEAT, Entrepreneur, Traveler, Interested in Technology,History, Startups and great conversation. Message me!
New York, Lisboa, Paris
Joined February 2008
- Tweets7.3K
- Following333
- Followers7.9K
- Likes1.5K
Antonio Evans retweeted
I’m super excited to announce that I’m joining @OpenAI as Chief Product Officer!
My entire career has been about working on big missions: connecting people and ideas at @Instagram and @Twitter, digitizing financial services at Libra/Novi, or most recently at @Planet, using space to advance global security, sustainability, and transparency.
OpenAI’s mission to build AGI to benefit all of humanity is perhaps the most important and transformative of them all.
Antonio Evans retweeted
Geoffrey Hinton says AI chatbots have sentience and subjective experience because there is no such thing as qualia
Antonio Evans retweeted
Just attended the kickoff event for Y Combinator W24 at OpenAI HQ with Sam Altman. Here are some key takeaways:
1. Startups need contextual optimization more than behavior optimization. Providing more information via RAG etc. can be more beneficial than fine-tuning. – at Palo Alto, CA
Antonio Evans retweeted
hope this is useful for people who spend the holidays thinking about what to go work on in 2024; always one of my favorite times of year
Antonio Evans retweeted
There exists no sentence in any language that conveys how happy I am:
Antonio Evans retweeted
We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Bret Taylor (Chair), Larry Summers, and Adam D'Angelo.
We are collaborating to figure out the details. Thank you so much for your patience through this.
Hey @ilyasut, bro, we're at a crossroads with AGI. Keeping it out of the corporate maze feels right for humanity, but man, the team's gotta be together too manage it. It's like choosing between heart 💗and mind🧠. Tough vibes, but I believe you'll find the path. #AGIEthics
Antonio Evans retweeted
Dear Ilya,
You're a great researcher but having witnessed first-hand how a billion-dollar venture ousted its CEO, here's my 2 cents on the matter:
When you're on the board of billion dollar company, even if it's an NGO, there are governance rules. You don't fire the CEO without the proper due diligence.
What does the proper due diligence look like:
1- Building a substantiated case supported by evidence.
2- Consulting all stakeholders - investors, management, researchers - to get alignment and majority support.
3- Discussing with major investors who committed billions before taking drastic action. You have an ethical obligation to do so.
What actually happened:
1- The CEO was dismissed abruptly over video call, without due process.
2- It seems that the Majority stakeholders were actually aligned with the ousted CEO, suggesting inadequate consultation.
3- Potential damage to reputation, valuation and industry trust wasn't fully considered.
What you ought to do:
Own the decision, but acknowledge mistakes in process.
Everyone involved in this mess should resign to demonstrate accountability and maybe help restore a bit of trust back.
And please don't stop your research, we all appreciate the value you've been contributing as a researcher.
Best.
Antonio Evans retweeted
i loved my time at openai. it was transformative for me personally, and hopefully the world a little bit. most of all i loved working with such talented people.
will have more to say about what’s next later.
🫡
Antonio Evans retweeted
Can GPT-4 teach a robot hand to do pen spinning tricks better than you do?
I'm excited to announce Eureka, an open-ended agent that designs reward functions for robot dexterity at super-human level. It’s like Voyager in the space of a physics simulator API!
Eureka bridges the gap between high-level reasoning (coding) and low-level motor control. It is a “hybrid-gradient architecture”: a black box, inference-only LLM instructs a white box, learnable neural network. The outer loop runs GPT-4 to refine the reward function (gradient-free), while the inner loop runs reinforcement learning to train a robot controller (gradient-based).
We are able to scale up Eureka thanks to IsaacGym, a GPU-accelerated physics simulator that speeds up reality by 1000x. On a benchmark suite of 29 tasks across 10 robots, Eureka rewards outperform expert human-written ones on 83% of the tasks by 52% improvement margin on average. We are surprised that Eureka is able to learn pen spinning tricks, which are very difficult even for CGI artists to animate frame by frame!
Eureka also enables a new form of in-context RLHF, which is able to incorporate a human operator’s feedback in natural language to steer and align the reward functions. It can serve as a powerful co-pilot for robot engineers to design sophisticated motor behaviors.
As usual, we open-source everything! Welcome you all to check out our video gallery and try the codebase today: eureka-research.github.io/
Paper: arxiv.org/abs/2310.12931
Code: github.com/eureka-research/E…
Deep dive with me: 🧵
