How to Invest in the Deep Learning of Autonomous Vehicles

On one side you have the engineering LiDAR approach, and on the other you have Tesla and startups like Wayve. Wayve is unique in that it’s using end-to-end deep learning to train autonomous driving software. It would make a great acquisition target for Apple or Amazon about now.

Projects like Google’s Waymo, Baidu, Cruise and Aurora are developing autonomous vehicles by throwing engineers at the problem, but what if Deep Learning could do the heavy lifting?

If you can wait to invest in Wayve when they go public, you’ll be investing in the AI of autonomous driving. You can see their website here. As Apple, Baidu and Huawei head to make EV cars, they will combine them with world-class autonomous driving programs. The same goes for the likes of GM, VM, Toyota and the pesky EV or AV startups.

The intersection of AVs and EVs might be one of the most significant technological advancements in the first half of the 21st century. Obviously investing at that intersection makes a lot of sense. Investing in a LiDAR winner might actually be a better bet than the ultra competitive EV sector, if you think about it.

Consolidation will occur very quickly in the EV, LiDAR and AV space and it will scale globally very quickly as well – faster than many assume today in 2021.

Wayve has only got $43 million in funding so far, which seems incredible considering how big it is as a team in 2021. Balderton Capital was the lead in their Series A. The LiDAR vs. heat-map debate is really heating up as Tesla promises to roll out its AV system called FSD. With Deep Learning can the smart car learn to drive from data using camera-first sensing without needing an HD-map?

Wayve is attempting to build the most adaptable AI driver, a Driving Intelligence that can learn to understand the world, yet adapt and scale intelligently to different driving domains: cities, vehicle platforms or mobility use-cases. It’s the classic humans (engineers) vs. AI problem and if it gets it right it could be an incredible company. Just as Amazon has invested in Rivian and Aurora, Wayve could be another feather in its cap too. Wayve however is using Azure, which is interesting.

AVs will have a huge total addressable market, and Waymo One and Baidu could be big winners, but what if they aren’t? Despite all this investment and many years of development, no one has yet been able to launch a commercial autonomous car service. What if a dark horse like Wayve makes a critical breakthrough? What if, contrary to most of the experts, Tesla actually gets it right?

Wayve believes an end-to-end machine learning system Is the “only scalable solution” for self-driving cars. Waymo says AVs are much more difficult than they first thought. It’s not clear how much progress Cruise has actually made, or how far ahead of them Baidu itself is. Let’s face it, robo-taxis and AVs are a bigger market than EVs, with all their hype and Tesla’s pathetic show-boating.

What Wayve is doing actually makes sense from an AI’s perspective. Instead of creating a platform and feeding it the rules of the road, their system uses imitation and reinforcement learning coupled with cameras and sensors to control the vehicle and follow a route entered in the navigation system. Think about it, that actually puts AI in the driver’s seat. Eventually this approach could win, if the AI is gets good enough at its job.

Can AI outperform engineers and billions of miles of real world experience? Wayve’s model-based deep reinforcement learning system gives it the ability to learn to drive like a human in new environments based on data either given or learned from past experience.

So what is Wayve’s input system? The input to their system is a video stream from 6 monocular cameras and some supporting sensory and ordinary sat nav information. Their neural network contains tens of millions of parameters and learns to regress a motion plan, which a controller is able to actuate on the vehicle. No fancy LiDAR here, it seems, absent.

Wayve was founded in 2017 by Amar Shah and Alex Kendall, two machine learning PhDs from the University of Cambridge. Wayve forgoes expensive lidar or data-intensive HD maps. Instead, the company’s technology is based on multi-layer neural networks and computer vision to predict the motion of objects in an urban environment.

Wayve believes they will get robo-taxis out first and be the first to deploy autonomous vehicles in 100 cities. Hey, Smart Car, can I give you a brain, please? Wayve have since deleted the blog post where they claim to be doing this. Not smart enough?

When Aurora acquired Uber’s self-driving program, you felt that Amazon in the end might be the surprise winner in the future of robo-delivery, logistics and the era of the robo-taxi. It’s highly likely Amazon will acquire both EV startup Rivian and AV startup Aurora. That would occur around the same time Apple makes its smart car in partnership with Foxconn in late 2026.

The future of EVs and AVs is not likely what you think, in 2021 it’s early days. Wayve’s deep learning system that is a learning-based system could end up being safer in unfamiliar situations than a rule-based system which would behave unpredictably in a situation it has not seen before.

Can end-to-end deep learning hold the key for the future of the AI driven smart car? If only DeepMind had been working in the real world doing something like this instead of doing AI research. Given the partnership of Wayve with Microsoft Azure, could Microsoft be the surprise ones to acquire them? It’s all very up in the air if you ask me but this is an AV startup to watch since their approach to solving the problem is so AI-centric.