This is an excerpt from Alex Heath SourcesA newsletter about artificial intelligence and the tech industry, distributed only to The Verge subscribers once a week.
Around the middle of last year, Pim de Witte began approaching several well-known artificial intelligence labs to see if they would be interested in using data from Medalits popular platform for cutting video games to train its agents.
Within weeks, it became clear that Medal's data was more valuable to the labs than he had expected. “We received a lot of acquisition offers very quickly,” he told me. (He declined to name names, but this was reported that OpenAI had offered $500 million.) “We were quite interested in them initially,” he said of the proposals, but “it was largely a result of us not understanding where we were sitting.”
He read Google DeepMind research work showing that game data can be used to teach AI how to navigate in a 3D environment. However, interest from artificial intelligence labs made him realize that his data from Medal, which receives about 2 billion video downloads a year from tens of thousands of video games, could be used to develop a unique basic model for bringing artificial intelligence into the real world.
“It's a pretty big bet.”
Today, Pim de Witte announced that Medal is creating a new artificial intelligence lab called General Intuition, which has raised a $133.7 million seed round. The money for the round largely came from Vinod Khosla, founder of Khosla Ventures and an early investor in OpenAI. Other investors include General Catalyst and Raine Group. Moritz Bayer-Lenz, who oversees Lightspeed's gaming investments, will also join the startup on a part-time basis as a member of the founding team.
Khosla believes that Shared Intuition could have the same impact in the field of agent AI that OpenAI has had in the way people use large language models. It's his firm's biggest seed since it backed OpenAI in 2018. “That’s a pretty big bet,” he told me. “They have a unique data set and a unique team.”
Unless you're immersed in the world of artificial intelligence, you probably haven't heard much about models of the world yet. This is a branch of research that trains AI to have human-like spatial understanding. The idea is that the robot could, for example, predict when a glass of water will spill if dropped from a table and grab it before it falls. On a more practical note, AI researchers are increasingly looking at world models as a way to train agents that can reliably create and interact with 3D space.
Among prominent AI leaders, Google DeepMind CEO Demis Hassabis has been the most vocal proponent of global models and their importance in achieving AGI. Google recently Humility Genie 3a model that creates a video game-like environment from the ground up as you move through it. There are also several startups working on similar models, including Fei-Fei Li's World Labs, which this week we released our own demo version model that generates interactive video in real time.
According to de Witte, the goal of Shared Intuition is to control any device that can be mapped to a keyboard and mouse or has an input circuit similar to a game controller. He expects the startup's first model to be used in search-and-rescue drones, but sees potential for applications in other areas, including humanoid robots and self-driving cars.
Just as LLMs were initially trained on text data from the Internet, de Witte believes that gaming environments will unlock the ability of AI to reliably predict correct actions in the physical world. “Games are essentially the only testable domain of spatiotemporal reasoning,” he explained. “You can tell a good action from a bad one, which is why it’s so valuable.”
However, it's a risky bet. The right technical path to develop models of the world is hotly debated in the AI industry, and as even Khosla pointed out to me, it's unclear what data will ultimately prove most valuable. Members of de Witte's early research group published notable study in this area, but the startup is still competing with better-funded giants such as Google. “Someone is going to win big in this market,” said Khosla, who told me he thinks this is an area where “several hundreds of billions and maybe even trillions of dollars worth of companies will be built.”
De Witte predicts that gaming companies will be a prime target for takeovers of artificial intelligence labs as interest in global models increases. His decision to found General Intuition was driven by the realization that Medal's data put him in a unique position to be more than just a data provider. However, he cautioned me that others might find it difficult to resist license checks and acquisition offers from large artificial intelligence labs.
“You're at an information disadvantage,” he said when I asked if he had advice for the games industry. “The better these models get, the less data they are likely to need.”
