Weekly Brief: AI Approach Gives Waabi a Driverless Edge

Self-driving car start-up Waabi came out of stealth mode with a bang last week.

The Toronto company announced a $83.5M Series A financing round led by Khosla Ventures. Uber and Aurora participated as investors as well, along with a host of leading artificial intelligence experts including Fei-Fei Li of Stanford University and Geoffrey Hinton and Sanja Fidler of the University of Toronto. The fundraise is among the largest Series A rounds ever in Canada.

The size of the investment is especially impressive given the challenging environment that self-driving start-ups face today. The past two years have witnessed one after another start-up get abandoned or subsumed into larger companies. Zoox got snatched up by Amazon at a bargain price of $1.3Bn, down from its former valuation of $3.2Bn. Uber folded its Advanced Technologies Group and had to pay Aurora $400M to take it off its balance sheet – in return, Uber got 26% of Aurora, but still. Lyft abandoned its Level 5 and sold it to Toyota for $550M in cash. Drive.ai is gone. Starsky is gone. Ike is gone. So many former darlings of the self-driving industry have turned to dust.

Yet, here comes Waabi out of stealth mode with a freshly loaded powder keg full of cash. What’s going on here? Look no further than Waabi’s founder and CEO, Raquel Urtasun. Urtasun isn’t just a woman working in an industry traditionally dominated from top to bottom by men. She’s widely regarded as a superstar with a once-in-a-generation mind for artificial intelligence. She spent years serving as Uber ATG’s chief scientist and head of the company’s Toronto unit. When Aurora acquired Uber ATG, a lot of eyebrows were raised when Aurora didn’t bring Urtasun onboard. Now it’s obvious why.

Urtasun believes that the traditional approach to engineering self-driving vehicles results in a software stack that doesn’t take full advantage of the power of AI and that requires complex and time-consuming manual tuning that makes scaling more costly and technically challenging. For anyone who disagrees with Urtasun, look no further than Waymo or Cruise, which collectively have logged millions of test miles yet still get thrown off by relatively simple edge cases. These lapses in turn make it more difficult, risky and expensive to expand their services.

Waabi claims that it does things differently, with a “breakthrough, AI-first approach that leverages deep learning, probabilistic inference and complex optimization to create software that is end-to-end trainable, interpretable and capable of very complex reasoning”. That’s an overdose of industry jargon but the gist is that Waabi creates a special, AI-powered software that can be trained on a closed loop simulator, which reduces the need to drive testing miles in the real world. The end result: a less capital-intensive solution that can be scaled more quickly without sacrificing safety.

Hence, the reason companies like Uber and Aurora are interested. Still, Waabi faces an uphill climb in the years ahead. The start-up’s focus is on self-driving trucks, where its competition is arguably strongest. Launching and scaling robo-taxi fleets in urban environments has proven vexing but getting autonomous trucks to operate on highways is a simpler task and everyone from Aptiv to TuSimple to Waymo is ramping up pilots quickly.

For example, last week Waymo revealed that it’s set to start moving freight for transportation giant J.B. Hunt. Waymo’s trucking service Waymo Via will transport cargo between Houston and Fort Worth, Texas. The self-driving trucks will be powered by Waymo’s Level 4 platform, the Waymo Driver, which in theory could handle the task alone but there will be a licensed truck driver, a software technician and an autonomous specialist on-board to monitor every mile of the test run. If things go smoothly, the two plan to scale the service.

I wouldn’t bet against Urtasun. Her start-up deserves the funding and hopefully will change the industry for the better, before it succumbs to an acquisition at a fraction of its forecasted valuation.


Leave a comment

Your email address will not be published. Required fields are marked *