General Intuition has raised another $220 million at a $6.2 billion valuation as the company starts making models available for training AI.
These are an entirely new class of foundation models, trained on billions of action-labeled videos, capable of acting in real time in never seen before environments, the company said.
Pim de Witte, CEO of General Intution, which is a sister company to Medal, said in a LinkedIn post that these are an entirely new class of foundation models, trained on billions of action-labeled videos, capable of acting in realtime in never seen before environments.
Leading foundation models in robotics and world models are trained on less than 1% of the action data we are capable of training on thanks to Medal, which is currently the fastest growing company in games, he said.
Medal is on track to three billion uploaded videos per year. Medal is the service where players upload clips of their gameplay.
“On any given day, we see more accidents in sim than across the entire USA,” de Witte said. “And at any moment, more people play with steering wheels than Waymo has cars on the road. This allows us to create General Agents and World Models, capable of handling hard to predict events.”
The company raised the round from Valor Equity Partners, Atreides, 776, Point72, Khosla Ventures, and General Catalyst. General Intuition has now raised more than $650 million. Back in June, General Intuition raised $320 million.
“We’ll be scaling up the team in New York and Europe. If you’re a top researcher or engineer, reach out,” de Witte said.
Why does this work? Humans play video games and (tele)-operate robots using the same inputs and outputs, namely vision & controller actions, he said.
“We train directly on the largest dataset of these, similarly to how Tesla trained FSD, but across more actions and environments,” de Witte said.
He added, “What emerges is a model capable of handling autonomy across a broad range of environments and possible actions, including in the real world. We’ve opened up the waitlist for partners who want to receive early access to the models.”
Partners can reach out for early access to models here. In a Twitter post, de Witte elaborated, saying, “leading foundation models in robotics and world models are trained on less than 1% of the action data we are capable of training on thanks to @Medal_TV, which is currently the fastest growing company in games. Medal is on track to 3 billion uploaded videos per year.
Why are the models so helpful? Humans play video games and (tele)-operate robots using the same inputs and outputs, namely vision & controller actions. We train directly on the largest dataset of these, similarly to how Tesla trained FSD, but across more actions and environments, de Witte said.
What emerges is a model capable of handling autonomy across a broad range of environments and possible actions, including in the real world.