Robotaxis drive miles just to get cleaned and charged; Aseon Labs wants to fix that; plus, General Intuition bets that video games can train AI agents for the real world

Speaker A: This is techcrunch. This episode is brought to you by Framer Summer is in full swing and for you, that might mean gearing up for big summer sales and events like if you're launching a July 4th blowout event. Stay on track with Framer. It's a complete website platform, not just a builder. With their AI agents, you can launch and update your site when you need to, fast and all in one place. The agents bring the speed and scale, you bring the taste, judgment and control. The pro site builder and design tools are perfect for getting every detail just the way you want. And here's the kicker. It's free. You only pay when you launch your site. Make your ideas go live in minutes with Framer. Learn how you can get more out of your site from a framer specialist or get started building for free today at ah framer framer.com tcstartups for 30% off a Framer Pro annual plan. That's framer.com tcstartups For 30% off framer.com tcstartups rules and restrictions may apply Take a stroll around San Francisco and it won't take long to spot an empty autonomous vehicle cruising the city streets, waiting to be hailed by a rider or heading off to a distant depot to be charged and cleaned. These deadhead miles, an industry term for miles driven without a paying passenger, are uh, one of the biggest barriers between robo taxi companies and profitability. Redwood City, California based uh, startup Asean Labs thinks it has a fix parking space sized automated pods that can be scattered throughout cities to inspect, clean and charge robo taxis. The company co founded by the team behind battery swapping startup Pushme, calls them robotic pit stops for the robotax taxi industry and the idea has caught the attention of investor Asean Labs has raised $10 million in a seed round led by Crane Venture Partners, Y Combinator, Uber Co founder Garrett Camp's venture firm expa, Robin Hood Ventures and Founders Capital also participated, along with a variety of angel investors. ASEAN Labs is still in the early stages. The seed funds will be used to build five prototypes of these pods, grow its six person robotics and engine team to about a dozen, and secure the real estate needed to build out its network, according to ASEAN Labs co founder and CEO George Calogeros. He said in order to reach economic parity with ride hailing, which is where we need to get with self driving cars and to stop really subsidizing the cost, you need the utilization to go up. You need the robo taxi in continuous operation during the entirety of the demand curve of the day ASEAN's pitch is that a network of distributed autonomous pods would slash deadhead miles and inevitably turn robo taxi services into profitable enterprises. Calogeros and co founder and COO Dan Keane come from outside the autonomous vehicle world, but they bring experience developing and scaling a hardware and real estate company. Calogeros uh worked as a mechanical design engineer at Bentley Motors and Tesla before he and Keane founded PushMe back in 2016 to build battery swapping infrastructure for micro mobility fleets. PUSHMI was building a battery swap network in Europe when it was acquired in January 2020 by Tier Mobility, Calogero said. The parallel that I'll draw is we were basically tasked by Softbank to put this across as many markets where it made sense for tier within a very short and compressed period of time. The playbook became how do we sprinkle the locations across the center of the city where it makes sense but at the same time make it easy to deploy as non permanent infrastructure? ASEAN Labs is applying the same thinking to autonomous vehicles. As they researched the industry, the pair visited AV depots where fleets of robo taxis are inspected, maintained, cleaned and charged. The cost of real estate often prompts companies to locate these depots outside the city center, which where most of the ride hailing activity occurs, he said. Depot infrastructure is the key requirement for the launch of a new city for any AV operator. And what happens in the depot right now, the operational backbone of Autonomy really is not fully baked. The founders settled on the idea of creating smaller, independently powered autonomous pods that could be dispersed throughout a city, but as importantly, they could also be moved as needed. The units, which include cameras to inspect vehicles and robotic arms to retrieve lost items and clean interiors, are considered temporary structures. That classification helps ASEAN Labs avoid a lengthy permitting process and allows the company to relocate units if a location underperforms. The units are designed to run on a propane generator for power or connect to an existing power source through partnerships with EV charging companies. They are meant to operate autonomously, although early versions will be staffed. According to Calogeros, ASEAN Labs isn't trying to tackle every edge case either. Instead, it leans on computer vision and AI, specifically Vision Language action models common in modern robotics to detect problems the pod should not try to solve. For example, if a camera detects melted chocolate on a back seat, the robotic arm stands down, since attempting to clean it could make the stain worse. Instead, the vehicle will be charged and dispatched directly to that company's central depot for a human to handle. ASEAN Labs has not signed contracts with any robo taxi companies yet, but Calogeros said there is widespread interest in the concept, saying pretty much everyone wants to try it. Before TechCrunch's Rebecca Balan entered General Intuition's R&D floor at UH, its New York office, the company's 31 year old co founder and CEO Pim DeWitt directed her attention to a monitor perched on a uh, standing desk. Someone appeared to be playing something like Fortnite, but it wasn't a person. Kent Rollins, the company's chief product officer, said, our agent has been playing for 100 hours straight. Before Rebecca could get absorbed in the spectacle of an AI navigating the game's virtual environment, she heard the electronic footsteps of a large quadrupedal robot approaching. DeWitt said the same brain powering the agent playing the game is powering the robot. Josh Duplantis, a data analyst carrying a laptop streaming a live feed from the robot's single camera piped up to explain that the bot's default mode was exploration. Relying on that camera, its singular eye, the giant bug like bot walked up to Rebecca, circled around her and continued into the office. It occasionally clipped the legs of chairs or bumped into an errant trash bin, much like a toddler who has not yet learned how her body relates to the world around it. Duplantis said it took just eight minutes of real world robotics data to fine tune an AI model for the Quadruped. What's more, that data was collected on the street, not inside the office where the bot was currently navigating itself. An agentic model that can generalize from gameplay to simulation to embodiment is General Intuition's raison d'. Etre. And that model's ability to figure out its place in the world has secured the backing of some pretty heavy hitters. On Thursday, General Intuition said it raised $320 million at a $2.3 billion valuation, confirming TechCrunch's previous reporting. The round brings General Intuition's total disclosed funding to 400,000 do $154 million after the $134 million round it raised at launch last October, the startup was spun out of DeWitt's other company, Metal, which allows gamers to upload and share video game clips. The hundreds of millions of hours of uploaded gameplay provided the initial data set to train General Intuition's model in spatial temporal reasoning or understanding how to move through space and time. But the key ingredient was not the gameplay footage. It was the action labels embedded, um, in those clips, records of exactly what buttons a UH player pressed and when most competitors, DeWitt says, are trying to infer actions from video alone, which he argues is insufficient, he said, we view this as just the next stage of future pre training. We have a single model that can respond to Fortnite information on the screen and and take action, but also to real world dynamics in a way that an LLM could never at one point, DeWitt sat Rebecca 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. Rebecca said that as often as she tests world models, she walked straight into a series of walls. In other demos that she tried, the agents that she controlled sometimes pass right through. But this one didn't. From the millions of hours of gameplay, it somehow learned that walls are walls, ladders are for scaling, and shadows lengthen as the sun moves. For M General Intuition, this world model isn't the product, it's the training environment, referred to as the gym. Internally, the company ultimately wants to sell the agentic model itself, and DeWitt argues that the action data embedded in gameplay is helps the model discern the self from the environment in a way that gives it a richer understanding of causality. Impressive Though General Intuition's technology appears in demos, the company isn't the only one trying to crack this problem. Moreover, getting such a model to hold up in the physical world at scale has not yet been done. Most approaches of this kind require enormous amounts of real world data that's gathered slowly and expensively. General Intuition's bet is that gameplay is a scalable shortcut. Its investors are OK with that bet as well. General Intuition's latest round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg and researchers at Google, DeepMind and MIT. The vast majority of the round will go towards scaling compute capacity. General Intuition has a deal with coreweave and plans to focus on pre training. The next version of the model, Slice, has been earmarked for making its API more broadly available by the end of summer. Vinod Khosla, whose firm led the round, says he was drawn to DeWitt's vision and the company's proprietary data position. Khosla told TechCrunch's Rebecca Balan, if you look at LLMs, when reasoning emerged, it was a quantum leap in world models. I think the quantum leap is the emergence of intuition in the AI, a human intuition like capability. The human action data and reaction data you have in games is the key part to the emergence of intuition. General Intuition is not the only company to notice that Metal's human action data is a key piece of the puzzle of building dynamic world models and general agents. Briana Martin, the startup's chief of staff, said the company was born in part after Metal turned down an acquisition offer from a major lab. There have been other offers since as well. DeWitt and his co founders are not interested in being acquired, and neither are the startup's investors looking for an exit just yet. The amount and quality of proprietary data General Intuition has via UH Metal is one of the reasons Khosla is convinced that the startup is a generational bet, not an M, and a target that it could become the backbone for generalized agents and world models in simulation and the real world. Part of that bet also involves trusting DeWitt's values. The entrepreneur spent three years working in the humanitarian space, including with Doctors Without Borders. As such, he has drawn a clear line for how General Intuition's tech will be used. No agents will be employed to harm humans, DeWitt said. We don't want to be an escalatory part of the system. Let's say I were to come out and say, uh, we're doing lethal autonomy. What do you think would happen in other countries? That limit on military use cases comes as Silicon Valley is growing ever more bullish on War, though DeWitt says he's happy for his models to be used for search and rescue missions. DeWitt is Dutch and much of his team is European, which shapes the company's identity. He says he brought on Martin in part due to her decision to publicly quit Palantir over its work with the U.S. immigration and Customs Enforcement. I don't know why Silicon Valley does what it does. There's a reason I'm not there. DeWitt's ethics don't simply limit what the models will not do. As a gamer who made $1.5 million by building and hosting a private Runesc server in his teens, DeWitt is also thinking about what happens to the people who get left behind by what AI models can do. General Intuition recently launched a platform called Nerve, a jobs marketplace that lets gamers earn money using their existing setups. Those who sign up start with data labeling and can eventually move toward robot teleoperation and other tasks. Metal's user base, DeWitt noted, is precisely the generation most exposed to AI driven displacement, and he wants them to have a stake in what's coming next. DeWitt wants general intuition to be an ecosystem enabler like Anthropic or OpenAI, a UM model provider that enables others to build on top of its technology. Today, the startup has a handful of customers in gaming, simulation and robotics, DeWitt said. We're not going to build a self driving car company. We're going to make it 10 times easier for the next person to build a self driving car company. The company says once it gets its API into more customers hands, it would be able to test its mettle with a variety of use cases, like testing a robot in a digital twin of a factory floor, powering a human like bot inside a gaming studio, or sending a quadruped to navigate hazardous environments. While a quadruped is the first physical embodiment that General Intuition has tried in the real world, it has also tried drones and other devices, including testing the model in driving games, DeWitt said. It works on anything that you can control using a game controller or a keyboard mouse. The possibility to build a data flywheel is one of its goals, DeWitt added. We'll pick customers where we can diversify the embodiments that this generalized foundation model is serving as the backbone for. So we're going to prioritize picking customers on whether they can offer real world data that's going to be interesting and useful to move the needle on research and if they'd have an agile internal team where we can be real embedded partners and learn from each other. Khosla said that General Intuition's proprietary data is what got it this far, and its ability to continue collecting data that no one else has will be essential. Especially because despite impressive demos, whether the simulation to real world transfer can hold at scale is an open question that nobody has fully answered yet. That's all for now. For the latest tech news, go to techcrunch.com.