Meta CEO Mark Zuckerberg hired the AI Avengers dream team to raise the full-fledged stakes for talent among the leading tech companies.
Zuckerberg announced on Monday the launch of Meta Superintelligence Labs. This is a group of star researchers who are tasked with poaching from AI competitors and building “personal tensions for all.”
Interest in the new unit surged after Openai CEO Sam Altman claimed that Meta was offering $100 million signature bonuses to recruits. But the team also has a glimpse into what Meta is doing with AI.
Meta is clearly interested in using multimodal AI. This means using AI to generate images, videos, and audio. We hired multiple people with expertise in this domain.
New recruits also show that Meta maintains a very close tab in Openai. Most of them are working on training the latest Openai models. (Meta is not the only big tech company obsessed with beating ChatGpt.)
Meta says he will disclose other employment later, so it's too early. Anyway, Employment Blitz has ordered Silicon Valley warnings. Whatever form, Meta's new team stands to shape the people who control the future of AI.
Meta did not comment on this article.
leader
Alexandre Wang According to Zuckerberg's memo, he leads the team as the best AI executive. At just 28 years old, Wang is already working together to LED scale AI, a startup that helps train the latest AI models. Meta recently invested $14 billion on a scale as part of a deal to hire Wang. Wang is very interested in AI progress. For example, while he was on a large scale, the startup helped create a particularly difficult test for AI systems called Humanity's Last Exam.
Natfriedman Zuck's memo says Will's Meta lab is working together with Wang. Friedman is best known as the former CEO of Github and as an AI investor who has supported startups such as Stripe and Prperxity. Friedman has also worked for Meta's AI Advisory Group since May 2024. He has also dabbled in other projects. This funds a programme to decipher ancient Roman scrolls that were burned by the Pompeii eruption.
Researchers and others
Trapid bored According to Meta's announcement, the former Openai research scientist co-created the company's leading O-series of AI models. Openai's O3 model is touted as Openai's most powerful “inference” model. Inference is a trend that has taken over AI over the past year, and involves AI chatbots embodying their “thinks” before answering questions.
Jiahui Yu According to his personal website, he was leading the Perception team at Openai, which works with multimodal AI, and the Perception team co-led with Gemini's multimodal efforts when he worked for Google. He also helps build some of Openai's latest models, says Meta.
shuchao bi We also worked on Openai's multimodal AI, which co-creates the GPT-4O audio mode. According to his profile page at Columbia University, he co-created YouTube shorts while working for Google.
Huiwen Chang According to Meta, she is a multimodal AI expert who helped launch image generation for Openai's GPT-4O model. Prior to that, she worked for Google and Adobe, according to her LinkedIn profile.
Ji Lin According to his personal website, he is a former Openai research scientist who specializes in multimodal and reasoning models. He is also the co-creator of some of Openai's latest AI models, says Meta.
horny He also worked for Openai and led a team focused on post-training AI models. “After training” means improving the performance of the AI model after the model itself has already been created.
Shengjia Zhao According to Meta, he is a co-creator of ChatGpt and previously led synthetic data at Openai. Synthetic data means using data generated by AI to make your AI model smarter. Another big AI trend as AI labs run out of materials and train them.
Johann Scharkwick I worked as a machine learning lead at Sesame. This is startup building software and hardware that allows you to naturally chat with people. According to the LinkedIn page, Schalkwyk previously worked on speech-related technology on Google.
Paysan Working for Google, I created the latest generation of AI models from Waymo, a subsidiary of Google's self-driving automobiles. According to Meta's announcement, Sun also worked on post-training and reasoning efforts for Gemini, a competitor of Google's ChatGPT.
Joel Pover We worked on building a reasoning system for Openai rival humanity. This means that a very popular AI system will allow enough data centers and other tools to run smoothly. Before joining humanity, Pobar worked for Meta (then Facebook) for about 10 years, with a major engineering team, his LinkedIn page showing.

