General Intuition: Bridging Virtual Gameplay and Real-World AI
As soon as I entered General Intuition’s R&D floor in its New York office, the company’s 31-year-old co-founder and CEO Pim de Witte drew my attention to a monitor perched on a standing desk. Someone appeared to be playing something like Fortnite. It wasn’t a person.
“Our agent played for 100 hours straight,” said Kent Rollins, the company’s chief product officer, beaming.
From Virtual Games to Real-World Robots
Before I could absorb the spectacle of an AI navigating the game’s virtual environment, I heard the electronic footsteps of a large quadruped robot approaching.
“The same brain that powers the agent playing the game powers the robot,” de Witte told me.
Josh Duplantis, a data analyst with a laptop streaming a live feed from the robot’s single camera, chimed in to explain that the robot’s default mode was “explore.”
Relying on this camera and its singular eye, the giant insect-like robot approached me, circled around me, and continued into the office. He would sometimes cut chair legs or bump into an errant trash can, much like a toddler who has not yet learned how his body relates to the world around him. Duplantis said it took just eight minutes of real robotic data to refine an AI model for the quadruped. Additionally, this data was collected on the street, not inside the office where the robot was currently navigating.
Investment and Vision
An agent model that can generalize from gameplay to simulation and embodiment is what General Intuition is all about. And the ability of this model to find its place in the world has gained the support of some heavyweights.
On Thursday, General Intuition announced that it had raised $320 million at a valuation of $2.3 billion, confirming TechCrunch’s previous report. This round brings General Intuition’s total disclosed funding to $454 million, following the $134 million round raised when it launched last October.
The startup spun off from De Witte’s other company, Medal, which lets gamers upload and share video game clips. The hundreds of millions of hours of gameplay uploaded provided the initial dataset for training General Intuition’s model in space-time reasoning – or understanding how to move through space and time.
But the key ingredient wasn’t the gameplay sequences; these were the action labels embedded in these clips: recordings of exactly which buttons a player pressed and when. Most competitors, de Witte says, try to infer actions from video alone, which he says is insufficient.
Training AI with Unique Data
“We see this as the next step in future pre-training,” de Witte said. “We have a unique model that can respond to on-screen Fortnite information and action, but also real-world dynamics in a way that an LLM never could.”
At one point, de Witte set me up with a laptop running General Intuition’s world model, a simulated environment generated frame by frame rather than rendered by a traditional game engine. As I often do when testing world models, I walked straight into a series of walls. In other demos I’ve tried, agents you control sometimes go straight through, but this one didn’t. Through millions of hours of play, he’s learned that walls are walls, ladders are for climbing, and shadows grow longer as the sun moves.
For General Intuition, this world model is not the product; this is the training environment (called “the gym” internally). The company ultimately wants to sell the agentic model itself, and de Witte says the action data embedded in gameplay helps the model discern “self” from “environment” in a way that gives it a richer understanding of causality.
Even though General Intuition’s technology appears impressive in demos, the company isn’t the only one trying to solve this problem. Furthermore, making such a model hold up in the physical world, on a large scale, has not yet been achieved. Most approaches of this type require enormous amounts of real data, collected slowly and expensively. General Intuition’s bet is that the gameplay is an evolutionary shortcut.
Its investors also accept this bet. General Intuition’s latest round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers from Google DeepMind and MIT.
The vast majority of the cycle will be devoted to increasing computing capacity. General Intuition has an agreement with CoreWeave and plans to focus on pre-training the next version of the model. A portion has been set aside to make its API more widely available by the end of summer.
Vinod Khosla, whose company led the round, says he was attracted to de Witte’s vision and the company’s proprietary stance on data.
“If you look at LLMs, when the reasoning emerged, it was a quantum leap,” Khosla told me in a telephone interview. “In world models, I think the quantum leap is the emergence of intuition in AI, an ability similar to human intuition. The data about human actions and reactions that you have in games is the key to the emergence of intuition.”
Visioning is a Generational Endeavor
General Intuition draws on data from Medal’s video game clips. Image credits:Medal.TV
General Intuition is not the only company to note that Medal’s human action data is a key piece of the puzzle of building dynamic world models and general agents. Brianna Martin, the startup’s chief of staff, said the company was born in part after Medal turned down an acquisition offer from a major lab. Since then, other offers have also been proposed.
De Witte and his co-founders, Eloi Alonso, Adam Jelley and Vincent Micheli, are not interested in an acquisition, nor are the startup’s investors looking for an exit at this time. The quantity and quality of proprietary data General Intuition has through Medal is one reason Khosla is convinced the startup is a generational bet and not an M&A target; that it could become the backbone of generalized agents and global models in simulation and in the real world.
“At this point it would be data acquisition, which is pretty uninteresting,” Khosla said.
Part of this gamble also involves trusting De Witte’s values.
The entrepreneur spent three years working in the humanitarian field, notably with Médecins sans frontières. As such, he drew a clear line on how General Intuition’s technology will be used: no agents will be employed to harm humans.
“We don’t want to play an escalatory role in the system,” de Witte said. “Let’s say I said, ‘We practice mortal autonomy.’ What do you think would happen in other countries?
This limit on military use cases comes as Silicon Valley is increasingly optimistic about war, although de Witte says he is happy that his models are being used for search and rescue missions.
De Witte is Dutch and much of his team is European, which shapes the company’s identity. He says he hired Martin in part because of his decision to publicly leave Palantir because of his work with U.S. Immigration and Customs Enforcement.
“I don’t know why Silicon Valley does what it does,” he said. “There’s a reason I’m not here.”
Future Directions and Ethical Considerations
De Witte’s ethics do not simply limit what models do not do. As a gamer who made $1.5 million creating and hosting a private RuneScape server during his teenage years, de Witte also thinks about what happens to people who are left behind by what AI models can do.
General Intuition recently launched a platform called Nerve, a job marketplace that allows players to earn money using their existing setups. Those who sign up start with data labeling and can eventually move into robot teleoperation and other tasks. Medal’s user base, de Witte noted, is precisely the generation most exposed to AI-driven displacement, and he wants them to have a stake in what comes next.
A Wheel of Data
De Witte wants General Intuition to be an ecosystem enabler, like Anthropic or OpenAI – a provider of models that allows others to build on its technology. Today, the startup has a handful of clients in gaming, simulation and robotics.
“We’re not going to create a self-driving car company,” de Witte said. “We’re going to make it 10 times easier for the next person to start a self-driving car company.”
The company says that once it gets its API into the hands of more customers, it will be able to test its mettle in a variety of use cases, like testing a robot in a factory’s digital twin, powering a human-like robot in a game studio, or sending a quadruped to navigate dangerous environments.
Although a quadruped is the first physical incarnation that General Intuition has tried in the real world, it has also tried drones and other devices, including testing the model in driving games.
“It works on anything you can control using a game controller or keyboard-mouse,” de Witte said.
The ability to build a data steering wheel is one of the goals.
“We will choose customers for whom we can diversify the embodiments on which this generalized basic model serves as the backbone,” de Witte said. “So we’re going to prioritize selecting clients based on their ability to offer real-world data that will be interesting and useful for advancing research. And whether they have an agile internal team where we can be true integrated partners and learn from each other.”
Khosla said General Intuition’s proprietary data is what got it to this point, and its ability to continue collecting data no one else has will be key. Especially since, despite impressive demonstrations, the question of whether the transfer of simulation to the real world can hold up on a large scale is an open question that no one has yet fully answered.
Correction: The title previously incorrectly stated how much General Intuition increased this turn. The error has been corrected.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
For more details, read the full article Here.
“`
