A 24-year-old AI researcher turned down $125 million offer from Meta, Zuckerberg doubled it up to Rs 21,791,975,000
According to a report by The New York Times, Deitke initially chose to focus on his AI startup Vercept instead of joining Meta's Superintelligence Lab. Two sources familiar with the talks said Zuckerberg's direct involvement changed that. The compensation was so significant that Deitke sought advice from trusted peers, many of whom urged him to take the deal. Eventually, he did.
This isn't just a story about money. It signals how aggressively Big Tech is recruiting AI talent as the race to dominate artificial intelligence intensifies.Matt Deitke's path to prominence didn't follow a typical corporate route. He began in a PhD programme in computer science at the University of Washington but left before completion to focus on real-world AI problems. That decision led him to the Allen Institute for AI (AI2) in Seattle, where he quickly stood out.
At AI2, Deitke led the creation of Molmo, a multimodal chatbot capable of processing not just text, but also images and audio. Unlike many language-based models, Molmo was designed to reason through visual and auditory data—an ambitious leap toward more human-like AI. The project was a success. His paper on the subject won an Outstanding Paper Award at NeurIPS 2022, one of the most respected conferences in the field. Out of more than 10,000 submissions, only a handful received the award.
This achievement established Deitke as a serious figure in the AI community, one whose work wasn't just experimental but functional and forward-looking.Molmo (short for Multimodal Language Model) broke away from the conventions of typical AI chatbots. It wasn't limited to written language. It could interpret photos, understand sound inputs, and respond in a way that accounted for context across different formats.The system was capable of spatial reasoning and interpreting real-time sensory input, giving it an edge over traditional large language models that rely purely on text. These features aligned closely with Meta's own AI vision, which aims to build systems that can not only talk but also understand the world in a more layered, perceptive way.Deitke's approach placed him at the forefront of AI research focused on perception and reasoning, two areas critical to advancing general-purpose intelligence.In late 2023, Deitke shifted focus again and co-founded Vercept, a startup aimed at building AI agents that act independently. Unlike typical AI tools that wait for human instructions, Vercept's agents were designed to identify goals, explore the internet, perform actions, and adjust to different digital environments.The idea was simple but bold: build AI that doesn't just respond, but thinks and acts. Vercept launched with just ten employees but quickly gained traction. The company raised $16.5 million in early funding. Among its high-profile backers was Eric Schmidt, former CEO of Google.With that, Deitke showed he wasn't only a researcher but an entrepreneur ready to challenge the industry from outside.Meta's Superintelligence Lab is a cornerstone of its AI ambitions. The company has already spent over $1 billion assembling what some call an "all-star roster." It includes researchers lured from rivals like OpenAI, Anthropic, Apple, and Google.One of the most notable recent hires was Ruoming Pang, who previously led Apple's AI models team. His reported compensation exceeded $200 million. In that context, Meta's $250 million offer to Deitke doesn't seem so out of line—it's part of a broader strategy to corner elite talent before competitors do.Deitke's combination of academic research, product vision, and startup experience made him a rare candidate. He had proven that he could not only theorise about the future of AI but actually build it.Matt Deitke's journey—from dropping out of a PhD to building multimodal systems and rejecting $125 million—has become a symbol of how the AI landscape is shifting. The brightest minds are not just employees anymore. They're independent thinkers, entrepreneurs, and public intellectuals with negotiating power.His eventual decision to join Meta suggests that the influence of Big Tech remains strong, but it's no longer unquestioned. Researchers like Deitke are now in a position to shape the direction of AI on their own terms.And Meta, like its rivals, knows it must offer more than just money. It has to offer vision, control, and the freedom to build what comes next.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
an hour ago
- Time of India
Qualifications that techies Mark Zuckerberg has hired personally at $100 million-plus pay packages have: Degree in ...
Meta CEO Mark Zuckerberg Mark Zuckerberg, CEO of Meta, is aggressively recruiting top AI talent for his new superintelligence lab, using a curated document known as 'The List' to target the industry's brightest minds. According to the Wall Street Journal, "The recruits on 'The List' typically have PhDs from elite schools like Berkeley and Carnegie Mellon. As a report in WSJ said, "The recruits on 'The List' typically have Ph.D.s from elite schools like Berkeley and Carnegie Mellon. They have experience at places like OpenAI in San Francisco and Google DeepMind in London. They are usually in their 20s and 30s—and they all know each other. They spend their days staring at screens to solve the kinds of inscrutable problems that require spectacular amounts of computing power." Zuckerberg has personally spearheaded the recruitment, diving into technical papers and strategizing with two Meta executives in a group chat dubbed 'Recruiting Party' to discuss hundreds of candidates and outreach methods, such as email, text, or WhatsApp. The List prioritizes candidates with PhDs in AI-related fields, experience at leading labs, and significant contributions to AI breakthroughs. Meta's recent hire, an algorithm design enthusiast, exemplifies the profile. The people who get notes from Zuckerberg reportedly have a few things in common: They need to know calculus, linear algebra and probability theory, like one of Meta's recent hires, who says he's fascinated by algorithm design. The AI community is tight-knit, with researchers forming Slack and Discord groups to share and discuss job offers, often leveraging competing offers to boost their value. Two sources familiar with these groups confirmed this practice. Meta has targeted dozens of OpenAI researchers, intensifying the Silicon Valley talent race. Leading Zuckerberg's superintelligence team is 28-year-old Alexandr Wang, a New Mexico native and son of Chinese immigrant physicists. Meta recently invested $14 billion in Wang's company, Scale AI, making him one of the most expensive hires in history. Meta has also reportedly made offers to dozens of researchers at OpenAI. This is said to have forced OPenAI to change its compensation structure. Not just this, the company's CTO also sent an open letter to employees asking them not to fall for Zuckerberg's offer. The company also shut down for almost a week. OpenAI's ex-CTO Mira Murati too has said that Mark Zuckerberg has also offered engineers in her company multi-million-dollar packages, but not even a single employee has so far accepted the offer. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Time of India
7 hours ago
- Time of India
New Instagram location sharing feature sparks privacy fears
Instagram users are warning about a new location sharing feature, fearing that the hugely popular app could be putting people in danger by revealing their whereabouts without their knowledge. The Meta-owned image sharing platform added an option on Wednesday which shares locations using an Instagram map, similar to a feature rival Snapchat has offered since 2017. Some users have since been shocked to discover that their location was being shared, viral posts have shown. "Mine was turned on and my home address was showing for all of my followers to see," Instagram user Lindsey Bell wrote in reply to a warning posted by "Bachelor" reality television personality Kelley Flanagan to her 300,000 TikTok followers. "Turned it off immediately once I knew but had me feeling absolutely sick about it." In a TikTok video, Flanagan called Instagram's new location sharing feature "dangerous" and gave step-by-step instructions on how to make sure it is turned off. Instagram chief Adam Mosseri fired off a post on Meta-owned Threads stressing that Instagram location sharing is off by default, meaning users need to opt in for it to be active. "Quick Friend Map clarification, your location will only be shared if you decide to share it, and if you do, it can only be shared with a limited group of people you choose," Mosseri wrote. "To start, location sharing is completely off." The feature was added as a way for friends to better connect with one another, sharing posts from "cool spots," Instagram said in a blog post. Users can be selective regarding who they share locations with, and can turn it off whenever they wish, according to Instagram. Wariness regarding whether Instagram is watching out for user privacy comes just a week after a federal jury in San Francisco sided with women who accused Meta of exploiting health data gathered by the Flo app, which tracks menstruation and efforts to get pregnant. A jury concluded that Meta used women's sensitive health data to better target money-making ads, according to law firm Labaton Keller Sucharow, which represented the plaintiffs. Evidence at trial showed Meta was aware it was getting confidential health data from the third-party app, and that some employees appeared to mock the nature of the information, the law firm contended. "This case was about more than just data -- it was about dignity, trust, and accountability," lead attorney Carol Villegas said in a blog post. Damages in the suit have yet to be determined.
Time of India
8 hours ago
- Time of India
Welcome to a World Where No One Needs to Learn a New Language
Live Events For centuries, language has been both a bridge and a barrier. It unites communities while isolating outsiders. But in 2025, we're no longer reliant on dusty phrasebooks or awkward hand gestures. Artificial intelligence, specifically through neural machine translation (NMT), has stepped up as the new global interpreter. Tools like Google Translate DeepL , Meta's SeamlessM4T, and OpenAI's Whisper aren't just translating—they're decoding the cultural DNA embedded in the AI translation revolution is more than an engineering achievement. It's a geopolitical realignment. It's a media shakeup. It's a cultural negotiation in real time and with real dictionary checks. Modern AI translation software is driven by massive language models (LLMs) that have been trained on billions of sentences. They don't merely decode syntax—they read context, tone, idioms, and meaning. They scan patterns in multilingual datasets and internalize the differences between, for example, Japanese sarcasm and British deadpan of those is DeepL, which has gained a cult-like following from professionals due to its smooth, human-sounding translations. Another one that's changing the game is Meta's SeamlessM4T—providing speech-to-text and speech-to-speech translation in almost 100 languages. It's not merely allowing you to read a foreign menu—it's enabling you to have an entire business meeting with no overlap of native languages at in multilingual goldmine India, AI is quickly becoming indispensable. With more than 122 principal languages and thousands of dialects, scalable translation is the key to digital inclusion. Open-source AI models such as Bhashini are working on platforms to enable Bharat to converse, read, and communicate in any translation isn't something; it's a paradigm. It's changing the way we form relationships, practice diplomacy, provide services, and narrate stories. It can unlock potential, give voice to, and redefine borders not only geographic but cognitive and emotional we can't afford to treat it as magic. Behind each translated sentence is a mesh of assumptions linguistic, cultural, and political. In order to construct a global society, we require more than machines that "understand" language. We require systems that honor future of translation isn't about saying the same words; it's about carrying the same meaning. And that's a problem AI can't solve on its own.
