A new startup called Wayve is now seeking to disrupt the autonomous vehicle industry with a focus entirely on machine learning. That actually flies directly in the face of the majority of efforts undertaken in that segment of the automotive market. So far, those have centered with an almost single-minded determination around the inclusion of more sensors and better mapping. Despite those endeavors, there have been a number of accidents and progress has arguably been difficult. The tenacious U.K.-based startup company Wayve, on the other hand, doesn’t think that’s a problem that can be quickly or easily solved via the abovementioned methods.
By placing simplicity and the learning capabilities of A.I. at its core, the company is effectively betting that the technologies involved in self-driving have just become too complicated. Wayve wants to build out a full driving software stack. What’s more, it wants that A.I. to be data-driven and to view the world more closely to how humans drivers do. That would allow it to adapt, the company says, from city to city without the need for new maps to be drawn up or a lot of new data. Of course, that’s still going to rely on sensors but it takes a step away from how other companies are approaching things. Instead, it puts the focus on not only detecting objects and people in the vicinity but also correctly identifying those and analyzing patterns in real-time. It’s an approach that puts A.I. back at the front and center. Better still, if successful, the advancement might be a lot safer than more constricted systems. Since the vehicle’s onboard computer would, in effect, be thinking through each situation at every layer, newly encountered scenarios wouldn’t necessarily result in errors. With current “rule-based” systems, new situations that aren’t pre-programmed are much more complicated.
On the surface, that approach seems less complicated and cleaner than those used by other companies. However, it remains to be seen whether or not the company can successfully upend the current direction of the autonomous vehicle industry. What’s more, it will likely take a substantial influx of resources and development to create an A.I. advanced enough to cope with driving tasks while still obeying laws of the road. With that said, there’s an argument to be made for Wayve’s proposed solution. Not only does the solution appear as though it would be safer in previously unencountered or unanticipated scenarios. The system would almost certainly require less hardware and, by proxy, fewer points of potential failure.