Zoox founder roars back with Twiggy-backed start-up

Tim Kentley Klay, who built autonomous vehicle company Zoox from nothing to a $US3.2 billion ($4.5 billion) valuation in just four years before being suddenly fired, has formed a new start-up to rival his former company with the backing of billionaire Andrew "Twiggy" Forrest.

The Australian entrepreneur had built Zoox into an industry-leading player in the rapidly advancing driverless vehicle space, where he was competing against a range of car industry incumbents as well as tech giants like Google, Uber, Tesla and Apple. He is attempting to leap ahead in the same market once more, with his new company HYPR.

Kentley Klay had raised about $US800 million in total funding, with a plan to create a ride hailing fleet, but was abruptly fired in August 2018, shortly after the completion of a $US500 million funding round led by Atlassian co-founder Mike Cannon-Brookes.

Zoox was then sold to Amazon in a relatively cut-price $US1.3 billion deal in June.

His new company will now pick up on Zoox's main goal, but with new autonomous vehicles Mr Kentley Klay says are based on more sophisticated artificial intelligence than any others in the market.

He tells The Australian Financial Review that HYPR's vehicles will use more advanced AI than Zoox, Google's Waymo, Uber and others already in the race to market. The software piloting the vehicles is in testing, and he is eager to leave "stealth mode" and get back into the fray.

"HYPR is a robotics artificial intelligence company, which I'm tremendously excited about. The vision to create ahead-of-the-curve autonomous mobility products,” Kentley Klay says.

Alongside Mr Forrest, HYPR is also being backed by US-based venture capital firm R7 Partners, with a $US10 million seed investment that already values the nascent company at $US200 million.

While HYPR’s purpose may sound very much like Zoox mark II, Mr Kentley Klay says it is the branch of AI that he is using that will make it a step change ahead of the current driverless vehicle makers.

Like the AlphaGo software, which learned to beat the world’s best human players at 2500 year-old Chinese board game Go, Mr Kentley Klay says HYPR’s vehicles will be driven by a type of AI based on reinforcement learning that effectively teaches itself.

The system performs actions in different environments and develops its own set of values to evaluate the success or failure, and thereby develops policies that vehicles will follow in future.

“One of the hardest things about autonomous mobility is the prediction piece; what's around me and what should I do? If you get that wrong, you have an incident and you're off the road, if you get it right you have a very big business on your hands,” Mr Kentley Klay says.

“So at HYPR, we are using reinforcement learning on test vehicles. It sounds very sci-fi, but they actually write their own code of how to drive … The belief is that if the AI develops these policies, it ultimately leads to a more robust system, that is high performance and with less compute power required at the same time.

“HYPR will be a next generation, integrated autonomous mobility company. We are developing the product, the vehicle and the artificial intelligence.”

Mr Kentley Klay says Zoox and other rivals like the General Motors-owned Cruise, Waymo, Uber and Tesla are using a “second generation” of AI, which combines classically coded systems with some deep learning, whereas HYPR represents a third generation he describes as an end-to-end learning system.

The system is based on vision and works on neural networks, is connected to the cloud and learns as it drives.