Big News in Autonomous Vehicles? Remains to Be Seen (but I can’t help being a little skeptical)
A self-driving vehicle that can learn from its experience? Such is what the (obscure) English startup, Wayve, is claiming. If what this company says turns out to be accurate, don’t expect its obscurity to last – this startup will be on the frontlines of autonomous vehicle innovation. Its name should be on the lips of everybody who touts the Ubers, Auroras (the autonomous vehicle offshoot of Amazon), Cruises (the GM autonomous vehicle branch), and Waymos (Google’s autonomous vehicle project) of the world.
What immediately strikes you when you’re talking about Wayve is not just its approach to the question of autonomous vehicle operation (as different from its competitors as motor oil is from water) but its gap in funding when compared with the largest companies on the landscape of self-driving cars (you can imagine the stacks of research dollars that these well-heeled companies are throwing at the conundrum). Wayve is surviving (according to Crunchbase) on just 300,000 British pounds (about $400,000 American dollars). Could a company with just $400,000 in funding really compete in (and even win if Wayve’s claims are accurate) the race to run a completely-autonomous vehicle effectively and safely on public roads, something that none of its deeply-pocketed, household-name, revered competitors have yet accomplished, something that none of the other small startups seeking to penetrate the same problem have managed to do either? Wayve is a relatively fresh-faced company, having come into existence in 2013 (but it wasn’t on anybody’s radar until Techcrunch brought the startup to our attention last year – at that time, Wayve had nothing much in the way of output to show for its troubles – we weren’t sure then why this company would even interest Techcrunch; apparently there was an inkling of what was to come).
The fact that not even the big boys, the 800-lbs gorillas in the space, have been able to develop a commercial vehicle that safely operates itself – after all of the dollars invested in specialized personnel, human engineers, and technology, and time devoted to developing such a vehicle (their efforts prove the difficulty of the endeavor) – this reality has spawned a surge of startups (Wayve among them) that are attempting to use artificial intelligence (AI) in place of complicated computer hardware and pre-programming methods to resolve the operational issues that are inhibiting the development of these autonomous road machines. Using AI as a solution is a necessity given the constraints on funding that most startups face (some of them will certainly garner plenty of large investments if and when their offerings pan out – given the landscape and field of this tantalizing industry – it’s the tantalizing possibilities of the industry that have attracted the likes of Google, Amazon, and General Motors). Those players, if honest, would concede that their success in the field has at best been limited.
So far self-driving vehicles have conquered highly-controlled situations in which adaptability to unforeseen occurrences (being dangerously cut off in traffic for instance or having to stop abruptly if somebody’s pet cat or dog or worse, somebody’s young child, was to venture into the street unexpectedly) and the navigation of uncharted territories (roads and terrain that the vehicle is encountering for the first time) are not on the table. Most experts consider Waymo to be tops in the field of autonomous vehicles, and the Waymo technology relies heavily on mapping and pre-programmed rules, and literally (in number) a bursting corral of sensors placed on, around, and within the vehicle. Without all of this the vehicle wouldn’t function – and this technical apparatus, required for operation, is both high in cost and heavy – leaving the machine inefficient and environmentally unfriendly as it must burn resource at an accelerated clip. Wayve’s first public demonstration of its emancipated vehicle (one that can adapt to conditions and, utilizing AI, will process information, rely on experience, and learn to navigate and drive safely the way a novice human operator would – this improvement by experience is a prospect both good and bad – remember how questionably you drove as a teenager, a novice driver? I cringe when I recall my own teenage driving; I thought I knew everything when I entered driver training because I’d already compiled a little experience behind the wheel – the first time I went out on the roads with an instructor I ended up driving on the wrong side of the street and the seasoned mentor said drily that while they do it that way over in England, I’d find out pretty fast it was a bad idea here) – Wayve’s first demonstration of its technology involved a Renault Twizy (a cute little twin-seat electric manufactured in Spain, seven feet in length and weighing only 450 kg (just under 1000 pounds); unavailable here in the states, the Twizy is categorized in Europe as a big-scale quadricycle) that relied for road navigation on onboard cameras and rudimentary phone-generated GPS directions. Presumably this modified Twizy was encountering the Cambridge roads it traversed for the first time (however by then it had undergone about 20 hours of training – again, kind of like a new teenage driver would). To sum up, experience helps the car drive. More experience helps the car to drive better. Eventually accumulated experience allows the car to navigate roads and traffic safely – and without human intervention. And artificial intelligence (AI) makes this possible. Employing AI reduces computer workload, dramatically lowering levels of complexity, cost, and ponderousness in the technology needed for autonomous operation).
This is a noble goal, but in practice the prototype Twizy exhibited a lot of the problems that a young driver might. It didn’t travel on the appropriate side of the road (the left-hand side in the UK). It ran stop signs and failed to yield the right of way at intersections where this was required. It traveled on lightly-used streets, surrounding traffic at a minimum. It never showed in the test that it could maneuver on crowded roadways or navigate effectively at even normal speeds. In addition, the Twizy’s compact size was ideal for getting between cars on England’s narrow streets. It isn’t clear if a larger vehicle making the same trip could have finished without sideswiping another car or instigating a head-on collision. However (according to Wayve) its Twizy showed what it showed relying on technological complexities equal to what you’d find on a contemporary laptop, a pretty impressive feat if true despite the navigational failures demonstrated by the vehicle in the test. Observers did see the Twizy handle some labyrinthian situations, however it did so without a lot of conviction (tentative is the word one observer used), as if it wasn’t completely sure it knew what it was supposed to do (again like an inexperienced human driver probably would).
The latest exhibition by Wayve is a little more complicated. If its newest technology works as advertised, Wayve could in fact end up transforming the driverless-vehicle industry. For one thing, Wayve is no longer using just a compact vehicle for testing purposes – rather it has outfitted a Jaguar I-PACE SUV (the very same vehicle that has been used by Waymo in some of its testing). The I-PACE is a five-passenger hybrid (employs both an internal combustion engine and battery power) with a gross weight of approaching 6000 pounds and with standard-sized measurements (instead of the compact Twizy’s specifications (substantial difference between machines of 1000 and 6000 pounds).
The SUV (according to Alex Kendall, Wayve’s co-founder and current Chief Technology Officer) boasts a world-first ability to drive on roads never before encountered, no pre-programmed HD maps of the terrain necessary, relying on none of the rules or sensors that characterize the operation of rival products. The Jaguar I-PACE does however require a safety driver to assume control if the vehicle confronts a situation or condition that it’s unfamiliar with (in other words, the SUV isn’t yet completely autonomous). Each episode of human intervention is supposed to teach the Jaguar to drive better and make it safer for navigation amongst pedestrians and ordinary thick traffic. Presumably this is the safest approach to vehicle self-driving available today, better than anything anybody else has developed or publicly tested (again according to Wayve’s Kendall). Anybody interested can view a video of a trip the Jaguar made through Cambridge at Jaguar (like the Twizy when it was tested, the Jaguar had never seen these roads and is said to be using common satellite navigation and basic unremarkable cameras as the catalysts for operation). One thing that seems apparent is that some of the timidness and operational uncertainty that spectators noted in Wayve’s test of the modified Renault quadricycle seem to have been overcome. Wayve’s less complicated approach is said to cost just 10% of what competitors spend on their ballyhooed technology – this is something that could help get self-driving technology to average consumers (a definite positive).
Now for my own skepticism. I (and others) would like to see some uninterested, unbiased third parties affirm the claims that Wayve has made. This hasn’t happened, but that doesn’t mean it won’t – the technology is still in its relatively early stages of development (apparently not available last year when Techcrunch profiled the company – so yes, it’s new). Newness entails technological bugs that must be addressed, again very common for any new technology. So far we only have Wayve’s assurances that its tests on the video-captured Cambridge roads, on other thoroughfares and streets that weren’t recorded for public consumption, occurred without the vehicle being familiar with them, without it being introduced to them before – also that the demonstration occurred without assistance from onboard maps and other preprogramming tools that competitors have relied on (there have been successful test runs with these competing systems – the big problem has been the consistency of operation, particularly the consistency of safe operation – operation that won’t routinely put pedestrians or other drivers at risk). There is also the fact that a human driver has to be present in the vehicle to intervene (granted that this intervention is supposed to become a teaching tool for the vehicle, qualifying it for increasingly complex and challenging self-operation, and therefore is a part of its makeup rather than a detriment proving it doesn’t work). However it doesn’t sound to me (nor does the video of the Jaguar’s path through Cambridge look it) much different from the self-driving mode in my amigo’s Tesla. The Tesla is capable of navigating itself through most basic road conditions, but it makes mistakes and certainly requires driver intervention – if the driver’s smart anyway – various Tesla crashes have proved that there are some idiot drivers out there among us). I question if the Wayve system is vastly different from Tesla’s. If it isn’t, then all of these problems plaguing competitors, those that Wayve claims to have addressed, in fact haven’t been adequately addressed at all.
I also haven’t seen near enough to fully believe in this technology given the difficulties in developing a fully-autonomous vehicle that have slowed down the likes of Google, Amazon, and GM, even after hundreds of millions of dollars (billions of dollars probably) having been flung at such problems – in an attempt by these mammoth companies to not only resolve the same nagging issues, but to resolve them before anybody else. Being first for anybody in an industry where your competitors are Amazon, Google, General Motors, and Uber would warrant bragging rights (to say nothing of the fortune that will come with those bragging rights).
I hope that I’m wrong in my skepticism about Wayve’s technology. I’d like to see it be real. I’d like to see self-driving cars on the road, sooner rather than later. I’m anticipating the day when that’s a reality, when I won’t need to worry about some dumbshit tailgating me at dangerously high speeds, cutting in and out of thick traffic in a haphazard manner, trying to text while behind the wheel, being impaired on the road by booze or some other mind-altering substance. Autonomous vehicles should also cut down significantly (won’t eliminate them altogether I’m afraid) the incidents of road rage that plague, even threaten us today (remember what happened to Joe McKnight?). One person who must believe in this technology (at least according to Wayve he’s among the company’s early-stage investors though the extent of his investment is unknown) is Zoubin Ghahramani, Uber’s Chief Scientist (not a bad person to have feeding dollars into your enterprise). So we’ll just have to stay tuned, keep our ears open, see what more happens. I can’t wait.