• Meta hires Shengjia Zhao, ChatGPT co-creator, as chief scientist of its Superintelligence Labs.
• Mark Zuckerberg is on a multibillion-dollar AI spending spree, which includes poaching talent.
• Other tech CEO have mixed opinions about Zuckerberg's recruitment approach.

Meta just escalated the AI talent war with OpenAI.

Shengjia Zhao, a co-creator of ChatGPT and former lead scientist at OpenAI, is joining Meta as chief scientist of its Superintelligence Labs.

CEO Mark Zuckerberg announced Zhao's appointment on Friday in a social media post, and called him a "pioneer" in the field who has already driven several major AI breakthroughs.

Zhao previously helped build GPT-4 and led synthetic data efforts at OpenAI. According to the post, Zhao will now work directly with Zuckerberg and Meta's newly appointed chief AI officer, Alexandr Wang, the founder and CEO of Scale AI.

The new hire comes during Zuckerberg's multibillion-dollar AI spending spree, including a $15 billion investment in Scale AI and the creation of Meta Superintelligence Labs, a new division focused on foundational models and next-gen research.

In addition to Zhao, the company has lured away the three researchers who built OpenAI's Zurich office — Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai — all of whom previously also worked at Google's DeepMind. The Superintelligence Labs team is now comprised of a lineup of names previously seen with OpenAI, Anthropic, and Google.

But the war for AI talent is far from over.

Databricks VP Naveen Rao likened the competition to "looking for LeBron James," estimating that fewer than 1,000 people worldwide can build frontier AI models.

Companies without the cash for massive pay packages are turning to hackathons and computing power as incentives. Perplexity CEO Aravind Srinivas said a Meta researcher he tried to poach told him to ask again when the company has "10,000 H100s."

AI tech workers have previously told Business Insider that Meta's Mark Zuckerberg has been emailing prospects directly and even hosting AI researchers at his home, while OpenAI CEO Sam Altman has made personal calls to potential hires.

Tech company executives have mixed feelings about Meta's poaching efforts.

"Meta right now are not at the frontier, maybe they'll they'll manage to get back on there," said Demis Hassabis, the CEO of Google DeepMind, on an episode of the "Lex Fridman Podcast," which aired on Friday.

"It's probably rational what they're doing from their perspective because they're behind and they need to do something," Hassabis added.

During a July 18 episode of the podcast "Uncapped with Jack Altman," OpenAI CEO Sam Altman criticised some of Meta's "giant offers" to his company's employees, and called the strategy "crazy."

"The degree to which they're focusing on money and not the work and not the mission," said Sam Altman. "I don't think that's going to set up a great culture."

Meta and OpenAI did not immediately respond to requests for comments.

Now that Meta has largely beaten an AI training copyright lawsuit raised by 13 book authors—including comedian Sarah Silverman and Pulitzer Prize-winning author Junot Diaz—the only matter left to settle in that case is whether Meta violated copyright laws by torrenting books used to train Llama models.

In an order that partly grants Meta's motion for summary judgment, judge Vince Chhabria confirmed that Meta and the authors would meet on July 11 to "discuss how to proceed on the plaintiffs’ separate claim that Meta unlawfully distributed their protected works during the torrenting process."

Chhabria's order suggested that authors may struggle to win this part of the fight, too, due to a lack of evidence, as there has not yet been much discovery on this issue that was raised so late in the case. But he also warned that Meta was wrong to argue its torrenting was completely "irrelevant" to whether its copying of books was fair use.

Read full article

Comments

On Saturday, Meta released its newest Llama 4 multimodal AI models in a surprise weekend move that caught some AI experts off guard. The announcement touted Llama 4 Scout and Llama 4 Maverick as major advancements, with Meta claiming top performance in their categories and an enormous 10 million token context window for Scout. But so far the open-weights models have received an initial mixed-to-negative reception from the AI community, highlighting a familiar tension between AI marketing and user experience.

"The vibes around llama 4 so far are decidedly mid," independent AI researcher Simon Willison told Ars Technica. Willison often checks the community pulse around open source and open weights AI releases in particular.

While Meta positions Llama 4 in competition with closed-model giants like OpenAI and Google, the company continues to use the term "open source" despite licensing restrictions that prevent truly open use. As we have noted in the past with previous Llama releases, "open weights" more accurately describes Meta's approach. Those who sign in and accept the license terms can download the two smaller Llama 4 models from Hugging Face or llama.com.

Read full article

Comments