On Tuesday 28 February, in the Banqueting Suite at one of London’s most iconic buildings, Tom Allebone-Webb, Head of Strategy & Innovation at Lloyd’s, introduced a packed audience of insurance professionals to Professor Paul Newman, Founder and CTO of Oxbotica.
The prestigious address in question was 1 Lime Street, the Richard Rogers-designed home of Lloyd’s of London, and the official title of the event was “The Future of Autonomy”.
The blurb promised an opportunity to “get up close with a zero occupancy all-electric self-driving vehicle which will revolutionise the goods and delivery market”, but there was much more to it than that.
The previous afternoon we spoke to Sam Tiltman, Sharing Economy and Mobility Leader for the UK & Ireland at Marsh, “the world’s leading insurance broker and risk advisor”, who described it as a call to action for a technological leap akin to the development of the internet.
“The combined impact of mobility as a service, electric vehicles and automation will be huge,” he said. “If autonomous vehicles deliver on their premise, they will significantly reduce risk, so if we don’t invest in this, then we, as an industry, will be disrupted.”
Insurance call to action
If any attendees were labouring under the impression that this is still the stuff of science fiction, the car parked outside the grand main entrance must have piqued their interest.
It was our first time seeing the record-breaking Applied EV vehicle for ourselves. However, regular readers will be familiar with its impressive radar vision, laser-based sensors and Oxbotica Driver System.
Passers-by were clearly more struck by what it doesn’t have – doors, windows, seats or a steering wheel. Last year it became the first autonomous vehicle to operate on UK public roads without a driver – a landmark achievement.
Both Tiltman and Applied EV CEO, Julian Broadbent, were also panellists for a lively Q&A, more of which later. First came Professor Newman’s presentation.
Dispensing with the faltering microphone, he spoke eloquently and with great passion for close to an hour. Opening with a crowd pleasing “insurance is awesome” message, he asserted that: “Insurance and autonomy are intertwined, because both will be everywhere.
“Since the days of the horse and cart we have persisted with the idea of one operator per vehicle. Now it can be ‘n’ operators per vehicle, and it will be insurers who decide what ‘n’ is.”
The four key questions that self-driving vehicles constantly ask, he said, are: Where am I? What is around me? What do I do? And what do I share?
He used the introduction of wolves to Yellowstone National Park as an analogy for the kind of “trophic cascade” that self-driving vehicles will bring about. For example, the energy-saving benefits of lights being turned off at automated ports – because autonomous vehicles don’t need them.
More pertinently, he continued: “Offline, the residual data from autonomy can be used to assess risk. Online, at a danger point, you can change behaviour. We call this insurance-in-the-loop.
“Assurance (and insurance) is king, so how do you test, verify and validate? We thought, let’s not waste our life scripting certain edge cases, let’s build AI to train the software. That’s Meta Driver.”
He finished with some eye-catching examples of scenarios encountered during real-world testing, and more challenging ‘deep fake’ examples of similar scenarios, created by tweaking variables such as timings, light levels and weather.
Self-driving vehicle manufacturing
There followed a short presentation by Broadbent, who explained his background in vehicle manufacturing and mission to design “cleaner, more efficient machines specifically for doing an autonomous job.”
On upscaling from the skateboard-like vehicle out front, through small cab-less delivery vehicles, to bigger trucks, he said the question was always: “What’s the next size up that we can make software defined?”
On public perception, he concluded: “We’re interested in how they interact and mainly people find them very dull, but there is a danger of a ‘scary clown’ problem.”
Next up was the aforementioned panel session, hosted by Allebone-Webb and featuring Newman, Broadbent and Tiltman, along with Chris Moore, of digital insurance specialist Apollo ibott 1971, and Rebecca Marsden, formerly of Apollo and now VP of Risk and Insurance at Oxbotica.
Opening the debate, Marsden said: “This is not just about risk management. The depth of data is transformative. It requires us to be collaborative.”
Tiltman agreed, encouraging a broader, more embedded role for insurers. “There is so much variance between different jurisdictions, we need to step up and help shape regulations,” he said.
Questions from the audience began with the reasonably blunt: “If you reduce risk sufficiently, do you need insurance at all?”
Moore responded: “Under our partnership with Oxbotica, we see them as a buyer of insurance today but a future distributor of insurance products or even a co-insurance partner in the future. This industry is currently very product focused, whereas we need to transition to being client and solution focussed. We have to break out of our silos and create a new product, an autonomy product.”
Further questions covered the possibility of a court requiring a developer to reveal “what’s inside the black box”, the possibility that unscrupulous actors might target self-driving cars in next-gen cash-for-crash scams, and the thorny issue of cyberattacks.
“The truth is the work is never finished, and insurers must be part of this digital solution – it will be ongoing, not something you can photocopy,” said Newman. “It will involve the sharing of best practice and keeping in mind why I got into this in the first place – safety. More parents should keep their kids, and more kids should keep their parents.”
It was a compelling point on which to end. For us though, the day was not quite done, as Newman generously found time to do a follow-up interview.
Paul Newman interview
We started with validation and the acceptability of simulation data.
PN: “Consider the combined experience of all vehicles in all places against the experience inside the skull of a 16-year-old just learning to drive. That’s all about risk, insurance and lifelong learning. Think about the hyperscalers’ access to data globally – it’s almost unfathomable.
“Just like computing, this is a technology we will be building on for all time. Of course, there has to be testing and statistics around actual vehicles operating in certain ways. But there comes a point when you have to ask: Can we augment that with something superhuman? Starting with real data is very important. We start with real images or real laser or radar data, and then massively exponentiate.”
We then tried to tease out what Oxbotica’s next big announcements might be.
PN: “We’re looking at energy, delivery, agriculture and construction; we’re very into 16-person shuttles in various cities, so that’s pretty much all vehicle types. It’s a pretty agnostic answer.” Well, can’t blame us for trying!
OK, final question: At the end of the event just now, people were talking about what happens in real time if the vehicle thinks it is off policy?
PN: “If the vehicle is off policy, then it could reasonably say it shouldn’t operate and could pause itself. But how interesting for a vehicle to be able to say, “Actually, I think I could be off policy, because I’m measuring increased risks”. To me there’s something glorious about that.
“Let’s not think about insurance as the net, but as part of the system that manages the risk. That’s really interesting and it’s going to cause a few recursions because the insurance itself is changing the behaviour, changing the risk.”
With motoring accounting for such a large percentage of insurance, that is “interesting” indeed.