This is an excerpt from Sources by Alex Heatha newsletter about artificial intelligence and the technology industry, is only distributed to The Verge subscribers once a week.
Around the middle of last year, Pim de Wit began reaching out to a number of prominent AI labs to see if they would be interested in using data from medalhis popular video game storytelling platform, to train their clients.
Within weeks, it became clear that Medal’s data was more valuable to laboratories than expected. “We received multiple takeover offers very quickly,” Lee said. (He refused to mention names, but… It has been reported That OpenAI offered $500 million.) “Initially, we were very interested in it, but that was mostly a result of us not understanding what we were sitting on,” he said of the offers.
He read Google DeepMind Research paper It shows that gaming data can be used to teach AI how to navigate in a 3D environment. However, AI Labs’ interest made him realize that his data from Medal, which receives nearly 2 billion video downloads annually from tens of thousands of video games, could be used to develop a unique foundational model for extending AI to the real world.
“It’s a very big bet.”
Pim de Witte announced today that Medal will create a new AI lab called General Intuition which has raised US$133.7 million. The money from the round comes primarily from Vinod Khosla, founder of Khosla Ventures and one of the early investors in OpenAI. Other investors include General Catalyst and the Raine Group. Moritz Baier-Lentz, who oversees gaming investments at Lightspeed, is joining the startup part-time as a founding member of the team.
Khosla believes that general intuition could be as influential in the field of AI agents as OpenAI has been in how people use large language models. It’s his company’s biggest seed check since it backed OpenAI in 2018. “It’s a very big bet,” he told me. “They have a unique data set and a unique team.”
Unless you’re immersed in the world of AI, you probably haven’t heard much about universal models yet. It is a branch of research that trains artificial intelligence to understand human-like spatial understanding. The idea is that the robot could, for example, predict when a glass of water will spill from the table and catch it before it falls. More practically, AI researchers are increasingly looking to world models as a way to train agents that can reliably create and interact with 3D space.
Among prominent AI leaders, Demis Hassabis, CEO of Google DeepMind, has been the most vocal advocate for universal models and their importance in achieving artificial general intelligence. Google recently demoed Genie 3, a model that creates a video game-like environment from scratch as you navigate through it. There are also a few startups working on similar models, including Fei-Fei Li’s World Labs, which This week he released his demo A template that creates an interactive video in real time.
For general intuition, the goal is to control any type of device that can be mapped to a keyboard and mouse or has a game console-like input system, according to De Witte. He expects the startup’s first model to be used in search and rescue drones, but sees the possibility of its application in other fields, including humanoid robots and self-driving cars.
Just as LLM students were initially trained on online textual data, de Wit believes that gaming environments will unleash the ability of AI to reliably predict the appropriate action to take in the physical world. “Games are basically the only verifiable domain of spatio-temporal thinking,” he explained. “You can separate the good work from the bad work, and that’s why it’s so valuable.”
However, it is a risky bet. The correct technical path for developing world models is hotly debated in the AI industry, and as Khosla pointed out to me, it is unclear what data will ultimately prove most valuable. Members of De Witte’s early research team I posted Notable research In this field, but the startup is still competing with better-funded giants such as Google. “Someone is going to win big in this market,” Khosla said, telling me he believes it’s an area where “companies worth hundreds of billions of dollars and maybe even trillions of dollars will be built.”
De Wit expects gaming companies to become prime targets for acquisitions of AI labs as interest in global models grows. His decision to start General Intuition was driven by the realization that with Medal data, he was uniquely positioned to be more than just a data supplier. However, he warned me that others may find it difficult to resist licensing checks and acquisition offers from large AI labs.
“You’re at a disadvantage in terms of information,” he said when I asked him if he had advice for the gaming industry. “The better these models get, the less data you will likely need.”