Sam Altman said Zuckerberg offered OpenAI employees up to $100 million signing bonus but none has taken it up
OpenAI CEO Sam Altman said Meta offered his employees signing bonus of up to $100 million and even larger annual compensation packages but none of OpenAI’s best people has left for Meta.
He thinks the reason why his employees are turning down the offers is because Meta is not great at innovation and there is a cultural risk of creating jobs that could become more about money than work.
Meta in talks to hire former GitHub CEO Nat Friedman and his NFDG partner, Daniel Gross, to join its AI team
Meta in talks to hire former GitHub CEO Nat Friedman and his NFDG partner, Daniel Gross, to join its AI team, the Information reported citing a person familiar with the matter.
NFDG is an AI-focused VC firm that not only invests between $1 million to $100 million in top AI startups, but also actively supports them via an accelerator and computing infrastructure.
Friedman is part of Meta’s Advisory Group, an external advisory council that advises the company on technology and product development.
Meta would be able to use a lot of the compute capacity they are building in the core business for the next three years, Susan Li said
In the “Cheeky Pint” podcast, Meta’s CFO Susan Li said that if the compute capacity that they are building will be less than the computed needed for inference, they would be able to use a large part of it in the core business and where they expect the core business to be in the next three years (min 19:00).
"To some extent, we still use a lot of this compute capacity in our core business and where we expect our core business to be three years from now. Frankly, we use more compute capacity in our core business. But this expansion is not infinite, right? The real question is, what happens if two years from now you have built so much compute capacity that you can’t foresee that the fallback plan can provide a reasonable ROI if the main project doesn’t come true? I think this is a question that all of us will learn in the next few years, " she said.
She said that the company’s focus is free cash flow (min 17:00). Li said she had to replace the “EBITDA hat” given to her by Zuckerberg with “free cash flow hat” because she realized that depreciation is becoming an important metric in their business. She also mentioned that she always questions Zuckerberg’s bets when they are budgeting to ensure that they can deliver the best ROI (min 9:00).
Li pointed out that Zuckerberg is good at giving feedback (timely, direct and respectful feedback) (min 6:00).
Assessment
I have often liked that Meta has a strong back up plan in case they won’t need all the compute capacity they are building. It’s also good to see Li having the interests of investors at heart, ensuring that Zuckerberg doesn’t get overly excited in so many bets.
Daniel Gross is also the CEO and co-founder of Safe Superintelligence, an AI startup company whose mission is to build a superintelligent AI that is safe by design, prioritizing alignment and control from inception — no products or services until that goal is met.
Safe Superintelligence was co-founded by former OpenAI’s co-founder Ilya Sutskever in June 2024.
It’s reported that Meta tried to acquire Safe Superintelligence (valued at $32 billion in April) at the beginning of the year and hire Ilya Sutskever, but he refused the offer.
According to insiders and former Meta employees interviewed by Bloomberg, Meta is facing reputational hurdles in its AI strategy due to the following;
Always shifting AI priorities.
Several insiders are disappointed with Zuckerberg’s gimmicky pursuits such as Snoop Dogg AI character while rivals are focused on creating AI that can meet or exceed human intelligence.
Zuckerberg poaches three open AI researchers to join his superintelligence team
Meta poaches three OpenAI researchers, Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai to join his superintelligence team, Wall Street Journal reported citing an OpenAI spokeswoman.
The three were working at OpneAI’s Zurich office since December 2024 but before then they were part of Google’s Deep mind.
Meta has poached four more notable AI researchers from OpenAI.
The four researchers include Jiahui Yu, Shuchao Bi, Shengjia Zhao and Hongyu Ren.
Jiahui Yu led OpenAI’s Perception team, which is responsible for making AI models like GPT process and understand the world beyond text, including images, video, and possibly other modalities.
Alexander Wang will be Meta’s Chief AI Officer and will co-lead the Superintelligence team with former Github CEO Nat Friedman
According to an internal memo, Meta’s Superintelligence team will be co-lead by former Scale AI CEO Alexander Wang and former Github CEO Nat Friedman, the Wired reported.
Wang will be the company’s Chief AI Officer while Friedman will focus on AI products and applied research.
The new organization, Meta Superintelligence Labs (MSL) will include all Meta Platforms foundations, product, FAIR teams and a new lab focused on developing the next generation of AI models.
Here are other new recruits including those I already reported above:
Meta tests AI sales agents and AI-powered voice support to WhatsApp Business
Meta is testing Business AI agents that recommend products, assist with sales, and follow up with customers via WhatsApp.
It’s also introducing the ability for large businesses to make and receive voice and video calls in WhatsApp, including support for use cases like telehealth, paving the way for AI-enabled voice interactions.
Meta hires Ruoming Pang, Apple’s top executive in charge of artificial intelligence
Ruoming Pang, Apple’s top executive in charge of artificial intelligence will join Meta Platforms, Bloomberg reported citing people familiar with the matter.
Pang who joined Apple from Google in 2021 has been running a roughly 100-person team responsible for the company’s LLMs.
Meta confirmed that it’s hiring Pang.
Meta will pay Pang tens of millions of dollars per year.
Outgoing Meta AI researcher blames the company’s poor AI performance on a culture of fear and a lack of direction
Tijmen Blankevoort, a departing Meta AI researcher, wrote a 2,000-word internal assessment that blames the company’s poor AI performance on a culture of fear and a lack of direction, according to The Information
“We are in a culture of fear,” he wrote. “Most do not enjoy being here,” he added. “And they don’t even know what our mission is.” It’s not just dysfunction – it’s a metastatic cancer that is affecting the entire organisation.”
I=6 Meta reportedly has no plans to open-source its forthcoming AI models
Zuckerberg told his new AI researchers that he doesn’t plan to outsource their upcoming AI models as they approach “superintelligence” threshold, Bloomberg reported.
I=8 Zuckerberg said Meta will spend hundreds of billions of dollars on AI infrastructure
Zuckerberg said in a Threads post that Meta will spend hundreds of billions of dollars on AI compute to build superintelligence.
“For our superintelligence effort, I’m focused on building the most elite and talent-dense team in the industry. We’re also going to invest hundreds of billions of dollars into compute to build superintelligence. We have the capital from our business to do this,” he wrote.
He added that they are on track to bring the world’s first 1 gigawatt-plus supercluster online in 2026.
“We’re actually building several multi-GW clusters,” Zuckerberg said. “We’re calling the first one Prometheus, and it’s coming online in '26. We’re also building Hyperion, which will be able to scale up to 5GW over several years. We’re building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan.”
Meta shares edged up 1% following Zuckerberg’s announcement.
Yeah, Zuckerberg mentioned this in an interview with The Information’s TITV.
“People say I want the fewest number of people reporting to me and the most GPUs and so having the most GPU per researcher is definitely a strategic advantage, not just for doing the work but for attracting the best people,” he said.
