For the past two a long time, Raquel Urtasun, founder and CEO of autonomous trucking startup Waabi, has been developing artificial intelligence systems that may reason like humans.
The AI pioneer was previously chief scientist at Uber ATG before launching Waabi in 2021. Waabi launched with an “AI first” approach to speed up the industrial deployment of autonomous vehicles, starting with long-haul trucks.
“If you can build systems that can actually do that, suddenly you need a lot less data,” Urtasun told TechCrunch. “You need a lot less computation. If you’ll be able to do the reasoning in an efficient way, you need not deploy fleets of vehicles around the world.”
Building an AV kit using artificial intelligence that sees the world as a human force and reacts in real time is something Tesla is attempting to do as a part of its vision-based approach to self-driving. The difference, beyond Waabi’s convenience in using lidar sensors, is that Tesla’s fully autonomous driving system uses “imitation learning” to learn drive. This requires Tesla to gather and analyze thousands and thousands of videos of real-world driving situations, which it uses to coach its artificial intelligence model.
Waabi Driver, on the other hand, has conducted most of its training, testing and validation using a closed-loop simulator called Waabi World, which routinely builds digital twins of the world based on data; performs sensor simulation in real time; creates scenarios for load testing of the Waabi driver; and teaches the Driver to learn from mistakes without human intervention.
In just 4 years, this simulator has helped Waabi launch industrial pilot programs (with a human driver in the front seat) in Texas, many of which are through a partnership with Uber Freight. Waabi World also enables the startup to realize its planned industrial launch of a fully autonomous solution in 2025.
But Waabi’s long-term mission is much greater than just trucks.
“This technology is extremely, extremely powerful,” said Urtasun, who spoke to TechCrunch via video interview with a white board filled with hieroglyphic formulas behind her. “He has an incredible ability to generalize, is very flexible and develops very quickly. This is something that we can expand to much more than just trucking in the future… It could be robotics. They can be humanoids or warehouse robots. This technology can solve each of these use cases.”
The promises of Waabi’s technology – which can first be used to scale autonomous trucking – allowed the startup to shut a $200 million Series B round led by existing investors Uber and Khosla Ventures. Strong strategic investors include Nvidia, Volvo Group Venture Capital, Porsche Automobil Holding SE, Scania Invest and Ingka Investments. This round brings Waabi’s total funding to $283.5 million.
The size of the round and the strength of its participants is particularly noteworthy given the hits the AV industry has had in recent years. In the trucking space alone, Embark Trucks closed, Waymo decided to halt its autonomous freight operations, and TuSimple closed its U.S. operations. Meanwhile, in the robotaxi space, Argo AI faced its own shutdown, Cruise lost its California operating permits following a major security incident, Motional laid off nearly half of its staff, and regulators are actively investigating Waymo and Zoox.
“The strongest companies are built when you raise funds during really difficult times, and the AV industry in particular has seen a lot of setbacks,” Urtasun said.
That said, AI-focused players in the second wave of autonomous vehicle startups have delivered impressive capital raises this yr. British company Wayve is also developing a self-learning, moderately than rules-based, autonomous driving system, and in May closed a $1.05 billion Series C led by SoftBank Group. In March, Applied Intuition raised $250 million at a $6 billion valuation to bring AI to automotive, defense, construction and agriculture.
“What is very clear today in the context of AV 1.0 is that it is very capital intensive and really slow to progress,” Urtasun said, noting that the robotics and autonomous vehicle industries are hampered by complex and brittle artificial intelligence systems. “And investors, I would say, are not too thrilled with that approach.”
However, investors today are excited about the promise of generative AI – a term that wasn’t exactly in vogue when Waabi launched, but nonetheless describes the system Urtasun and her team created. Urtasun claims that Waabi is a next-generation genAI that might be deployed in the physical world. Unlike popular contemporary language-based genAI models akin to OpenAI’s ChatGPT, Waabi has found out create such systems without relying on massive datasets, large language models, and all the computing power that comes with them.
Urtasun claims that the Waabi driver has a remarkable ability to generalize. So as a substitute of coaching the system on every possible data point that has ever existed or could ever exist, the system can learn from a few examples and deal with the unknown in a protected way.
“It was by design. We’ve built these systems that may perceive the world, create abstractions of the world, and then based on those abstractions consider, “What might happen if I do this?” – Urtasun said.
This more human and reasoning-based approach is much more scalable and capital efficient, says Urtasun. This is also essential for checking safety-critical systems that operate at the edge; you don’t need a system that takes a few seconds to reply or you will crash the vehicle, she said. Waabi announced a partnership introducing Nvidia’s Drive Thor solution for autonomous trucks, which can give the startup access to automotive-grade computing power at scale.
On the road, it looks as if the Waabi driver understands that there is something solid in front of him and that he should drive rigorously. He may not know what this thing is, but he’ll know avoid it. Urtasun also said that the driver was capable of predict how other road users would behave without having to be trained in various specific cases.
“It understands everything without telling the system about the concept of objects, how they move in the world, that different things move differently, about occlusion, about uncertainty and how to behave when it rains a lot,” Urtasun said. “All these things are learned automatically. And because he is now exposed to driving scenarios, he is learning all these capabilities.”
She noted that Waabi’s simplified, single architecture may very well be applied to other autonomy use cases.
“If you expose him to interactions in a warehouse, picking things up and dropping things, he can learn it without a problem,” she said. “You can expose him to multiple use cases and he can learn to use all these skills at once. There is no limit to what he can do.”