Catching autonomous taxis will cost less than half of what it does to drive your personal car and a tenth of riding in conventional taxis. This is what will push wide-scale adoption of the technology, and currently, some of the main players overcoming the final challenges to self-driving cars are Tesla, Waymo, Baidu, and Wayve.ai. Today we invite Alex Kendall, Co-Founder and CEO at Wayve.ai onto the show to speak about how the company is differentiating itself and solving some of the problems to wide adoption of self-driving cars. The private company Wayve.ai aims to build scalable, adaptable robotics for learning algorithms for self-driving cars. They have made a bold move in committing to a full end-to-end deep learning approach, meaning they use machine learning to optimize driving algorithms right from the input to the output. Alex begins by explaining how this approach differs from competitor systems and presents a higher chance of finding a solution that will be able to handle any environment it is presented with. Our conversation with Alex then touches on some of the technical challenges he and his team are facing, covering topics around system interpretability, labeling, complexity, measuring progress without a static trained data set, and more. We then zoom in on some of the challenges in the area of commercialization, hearing Alex talk about the regulatory sphere, data sharing for training learning systems, and other relevant topics. Tune in for a rigorous interview that paints an accurate picture of the landscape of self-driving cars from one of the main players who is driving research forward.

Key Points From This Episode:

• Introducing Alex Kendall and how machine learning can serve society via self-driving cars.
• Arguments for the adoption of autonomous vehicles and who will manufacture them.
• The technology that makes Wayve.ai unique compared to Tesla, Waymo, and Baidu.
• A deeper dive into the approach at Wayve that utilizes end-to-end machine learning.
• Limitations and challenges in the sphere of deep learning relating to interpretability.
• Some of the challenges and possible solutions in areas of data and labeling.
• Wayve’s approach that measures progress without having to use a static trained test set.
• Challenges to the validation of new self-driving systems and how they might be overcome.
• The path to commercialization at Wayve.
• How Wayve intends to optimize capabilities in different environments and vehicle sizes.
• Wayve’s approach to onboard compute.
• How Wayve is handling the issue of latency using a basic inference system.
• The approach to handling regulations at Wayve: The evidence-driven safety case.
• Why self-driven public transport also falls into the purview at Wayve.
• How the issue of data sharing and access to data between competitors is progressing.
• Competitors Alex admires and his thoughts regarding exponential adoption.
• Assisted driving models versus full autonomy: Alex’s thoughts on adopting the former.
• The connection between intelligence and movement.
• Whether the neighborhood electric vehicle space is relevant to Wayve’s mission.
• Thoughts on how different car models will benefit needs as they evolve.

Tweetables:

“By learning end-to-end, it enables us to achieve a level of efficiency that you can't obtain with the traditional rules or LiDAR or an HD map-based approach that many of the incumbents use.” — @alexgkendall [0:05:49]

“I would highlight data as being a really important problem for us to solve and a big bottleneck. In particular, it's not the quantity of data, it's the quality and the distribution of data.” — @alexgkendall [0:13:31]


More FYI Podcasts: https://ark-invest.com/podcasts/

To learn more about ARK: https://ark-invest.com/​​

For more updates, follow us on:
Twitter: https://arkinv.st/3corDiY​​
LinkedIn: http://arkinv.st/1f7AiVX​​
Facebook: http://arkinv.st/3r4EDOU​​
Instagram: http://arkinv.st/39td8bO​​

Disclosure: http://arkinv.st/39rzF94​​