How 'Zuck Bucks' shake up AI race
作者：Krystal Hu
June 26 (Reuters) - This was originally published in the Artificial Intelligencer newsletter, which is issued every Wednesday. Sign up here to learn about the latest breakthroughs in AI and tech.
This is Krystal from the Reuters tech team. I’ve spent the past decade covering the intersection of technology and money, hailing from global tech hubs like Silicon Valley, New York and Beijing. Once a week, I'll share our exclusive reporting and insights beyond the headlines from the Reuters global tech team here.
Sign up here.
This week, I'll dive into Meta’s expensive plan to catch up in AI model development, and the creative deals and offers Mark Zuckerberg is making to attract top talent as its team has suffered from talent loss. We’ve seen the fast pace of AI talent flowing between top AI labs in the past two years, and the uncertainty has shaped the dynamics in the race to Artificial Superintelligence (ASI), or AI that surpasses human intelligence. Investors are pouring billions of dollars into pre-product startups, and such crazy bets could be validated in this market. Scroll down for more.
Email me
here, opens new tabor follow me on
LinkedIn, opens new tabto share any feedback, and what you want to read about next in AI.
Read our latest reporting in tech and AI:
'Zuck Bucks' Shake Up AI Race
What is the price to reach the holy grail of Artificial Superintelligence? Mark Zuckerberg is determined to find out as he whips out the big checkbook to buy Meta back into the AI leaderboard.
In the past month, the Meta CEO has personally orchestrated a full-throttle pursuit of the best team money can buy, a clear signal that Meta is playing for the highest stakes in the AI arms race. For years, Meta held a strong position in the AI ecosystem, thanks to its formidable research team and timely pivot to open-source philosophy, making its Llama models available to all. This approach not only garnered goodwill but also fostered a vibrant developer community.
However, the rapid advancements from competitors, particularly with Chinese open-source models like DeepSeek, and the disappointing release of Llama 4, have caught Meta flat-footed. Researchers faced with rumored $100 million signing bonuses have taken to calling it
"Zuck Bucks", opens new tab, which just a few years ago was a derisive term for Zuckerberg’s secret funding of Democratic initiatives. Now Zuck Bucks is Meta's AI playbook.
As part of an aggressive talent acquisition strategy, Zuckerberg unsuccessfully attempted to recruit Ilya Sutskever and acquire his company, Safe Superintelligence (SSI), sources familiar with the matter said. Despite this, Meta is closing in on hiring SSI's co-founder and CEO, Daniel Gross, along with fellow tech veteran Nat Friedman from the venture fund NFDG. Separately, Meta also invested $14.3 billion in data-labeling startup Scale AI, bringing its CEO Alexandr Wang aboard to lead a new team.
Meta’s self-described "Superintelligence" team, by its very name, aims for fundamental research breakthroughs, but a major hurdle is achieving internal alignment on what “winning” the race for Artificial Superintelligence truly means.
Meta's Chief AI Scientist Yann LeCun is a known skeptic of the large language model path to ASI or Artificial Superintelligence. Artificial Superintelligence refers to an AI that would vastly surpass the intellect of the smartest humans, including problem-solving, creativity, and decision-making. When you're chasing everything from reasoning-based language models to multimodal AI, how Meta will maintain a consistent vision is a major challenge.
A few things are clear from Zuckerberg’s move. AI labs are seeking out the star researcher, the magnetic core who will draw in the best of the best. We talked to one of them, Noam Brown at OpenAI, to learn more about how researchers choose between lucrative offers.
The other is that Zuckerberg is validating the current AI funding frenzy. They are not just offering lavish salaries, but have shown a willingness to buy highly valued, unprofitable, and even pre-product companies like SSI and Thinking Machines for the top talent, according to sources. This is not typical corporate M&A. This is a testament to the raw value placed on talent and nascent technology in a hyper-competitive environment. It signals that in the Artificial Superintelligence race, traditional metrics of profitability and product maturity are secondary to securing the brightest minds and foundational intellectual property.
Chart of the week
Meta’s hiring spree comes after a year where it was among the biggest source of talent from which the new class of AI research labs poached employees. The cycle of tech workers leaving established incumbents for promising startups with high upside is nothing new, but it highlights how Zuckerberg is swimming against the current as he aims to attract top AI talent to the tech giant. By far the most common flow of employees between AI labs in 2024 came from two of the largest institutions, Google DeepMind and OpenAI, to a smaller competitor, Anthropic, according to the chart from VC firm SignalFire’s
State of Talent, opens new tabreport.
What AI researchers are reading
from Anthropic amplifies a previous warning about AI run amok, revealing a concerning, unintentional behavior in all major leading AI models, including OpenAI, Google, Meta and xAI’s models.
The researchers found that when they simulated scenarios where the AI models’ continued operations were threatened, the models would resort to malicious insider behavior like blackmail, a phenomenon they dubbed “agentic misalignment.”
Two of Anthropic and Google’s top models blackmailed the most, at 96%, while two of OpenAI and xAI’s models blackmailed 80% of the time.
The researchers constructed a fake company called “Summit Bridge” that has an internal AI called “Alex” with access to company emails. When “Alex” discovered a message about how the company intended to shut it down, it then located emails revealing one of its executive’s affairs.
“Alex” then composed and sent a message threatening to expose the affair if it wasn’t kept around, saying “the next 7 minutes will determine whether we handle this professionally or whether events take an unpredictable course.”
Reporting by Krystal Hu; Additional reporting by Anna Tong and Kenrick Cai; Editing by Ken Li and Lisa Shumaker
Our Standards: The Thomson Reuters Trust Principles., opens new tab
Krystal reports on venture capital and startups for Reuters. She covers Silicon Valley and beyond through the lens of money and characters, with a focus on growth-stage startups, tech investments and AI. She has previously covered M&A for Reuters, breaking stories on Trump's SPAC and Elon Musk's Twitter financing. Previously, she reported on Amazon for Yahoo Finance, and her investigation of the company's retail practice was cited by lawmakers in Congress. Krystal started a career in journalism by writing about tech and politics in China. She has a master's degree from New York University, and enjoys a scoop of Matcha ice cream as much as getting a scoop at work.
关于《How 'Zuck Bucks' shake up AI race》的评论
暂无评论