For our hot new CleanTech Talk podcast interview series, I recently sat down with ARK Invest Analyst Tasha Keeney to discuss various aspects of autonomous vehicles, Tesla, Tesla, and Tesla. Tasha is focused on autonomous cars and 3D printing in her position at ARK Invest. She and ARK Invest CEO Cathie Wood recently hosted Tesla CEO Elon Musk in the studio for their new podcast series, so we built off of that chat and I also brought in several key autonomy topics I’m always eager to learn more about.
The conversation was approximately 45 minutes long, so to cut it down into more manageable portions, I’ve split it into two episodes. This is the first episode, and the second will be published tomorrow. Listen to the discussion on your favorite podcast platform (options linked below) or via this embedded player:
To kick off, I asked Tasha about her background and how it fit into ARK Invest. She was previously a management consultant, mostly in the auto industry on the supply chain side of the story. We slid into a short chat about the autonomous vehicle industry as a whole and suppliers in the industry before making our way to Tesla and the large team of experts at the company who are focused on advancing Tesla Autopilot on a daily basis, and who meet with Elon on a weekly basis.
I then brought up my favorite CleanTechnica article on Tesla autonomous driving, one written by Mike Barnard in 2015 and republished last year when Tesla hit 1 billion miles on Autopilot. I highly recommend reading that article before or after listening to the podcast. Tasha took that summary and got more technical, explaining part of why they are bullish about Tesla’s autonomy approach.
I also touched on a Navigant Research report that puts Tesla near the bottom of the list for self-driving vehicle strategy and development. I had found those conclusions to be confusing, but also had a general hunch why they were what they were, so I was curious to hear Tasha’s take on that. She focused on the point that Tesla doesn’t use prototype vehicles, which may confuse analysts, and that there’s perhaps over-reliance on California disengagement reports (which are actually not standardized and are self-reported).
Another top topic of curiosity for me for years has been exactly how and how much Tesla uses “ghost driving” or “shadow mode” to improve Autopilot and eventual Full Self Driving. Clearly, Tesla’s approach involves learning from drivers who have Autopilot on and then disengage for some reason, but I also wonder how much Tesla’s software is learning how to drive simply from monitoring what the human drivers do without Autopilot activated. Tasha’s take on this from talking to Elon is that the Autopilot disengagements are particularly important but that the system must be learning to some degree from the shadow mode testing. (This is really a topic it would be cool to talk about with Elon or one of his Autopilot team members, as there’s still plenty of mystery here.)
Lex Fridman at MIT has been tracking Tesla Autopilot miles and last we heard had the figure just over 1 billion. Tasha has built on that work for ARK Invest to come to the conclusion that Tesla vehicles with “Autopilot hardware 2” have driven 10 billion miles, better capturing or quantifying the miles that are potentially involved in ghost/shadow learning.
Getting to the topic of full self driving, Tasha touches on Elon’s comment on the ARK Invest podcast that he and the crew think Tesla vehicles “will be feature complete — full self-driving — this year,” and could be ready for drivers to go to sleep behind the wheel a year later — but that depends mostly on regulators. Tasha said that is probably an aggressive timeline, but considering it’s coming from Elon, how could it not be? Going back to the ARK Invest interview with Elon, these were his words on that matter: “My guess as to when we would think it is safe for somebody to essentially fall asleep and wake up at their destination? Probably towards the end of next year. That is when I think it would be safe enough for that.” Again, though, it’s up to regulators when that is permitted, so Tasha and I discussed that topic of regulators/regulations for a few minutes.
We then briefly talked flying taxis and electric vertical takeoff and landing (eVTOL) aircraft — fun topics, but that will be a topic for another analyst and podcast discussion.
We ended the first part of this two-part discussion by chatting about the processing power needed in autonomous vehicles, the importance of that for over-the-air software updates, and Tesla’s overall hardware and software leadership. Highlighting this challenge for conventional, established automakers, Tasha had this to say:
“We call it here at ARK the sort of ‘old DNA’ issue — when innovation happens that you can sort of be caught flatfooted if you’re very stuck in your ways. I think that’s a huge problem for the traditional autos. And same thing with autonomy.
“If they were to do this, they have an even larger fleet than Tesla, right? They could get this data. But Tesla started years ahead of them. So, we think it’s possible they could just totally run away with this opportunity.”
As part of this whole matter, Tasha and I talked about a presentation from George Hotz (aka geohot) and a brief Q&A I had with him in which he emphasized Tesla was so far ahead on software in part because the company was so far ahead on hardware. For more on that, I recommend this CleanTechnica exclusive: Geohot: Tesla Autopilot = Apple iOS, Comma.ai = Android.
The second part of our interview with Tasha Keeney will be published tomorrow here on CleanTechnica, including on the CleanTech Talk portion of our site. In the meantime, I recommend this article and podcast for more on these topics: Elon Musk: Full Self-Driving Teslas This Year, “Unequivocal” Tesla Autopilot Improves Safety.
Interested in buying a Tesla Model 3, Model S, or Model X? Need a referral code to get 1,000 miles of free Supercharging? Use ours: http://ts.la/tomasz7234 (or not).