Tesla – King Canute

Elon Musk is becoming a big problem.

• Outside of a chronic inability to scale, Tesla’s biggest problem is rapidly becoming its mercurial CEO where I think that his views on AI are causing friction and pushing the company in the wrong direction.
• Management of Tesla’s autopilot activity has changed hands multiple times in the last few years (see here) and now it appears that following the removal of Stuart Bowers by Elon Musk in early May this year, a further 10 or 11 key personnel have also left the company (see here).
• This kind of upheaval is never good for progress, and I think it is being made a lot worse by Elon Musk’s unrealistic expectations of what AI is capable of.
• Autonomous driving is recognised as a difficult AI problem but just how difficult is only beginning to come to light now.
• This is why we saw Waymo struggle in Arizona in Q4 18 and why everyone has subsequently massively pulled back their expectations of when this will be a commercial reality.
• Almost everyone that is except Elon Musk and by default Tesla.
• Tesla claims that it is miles ahead of everyone when it comes to autonomous driving, but at the same time, it declines to provide any data.
• This view of its superiority is based on the notion that it has solved the problem of machine vision.
• Humans can drive to a standard of 1 fatality per 100m miles driven using just their eyes and ears, leading Tesla to believe that machines can do the same using 8 cameras and radar for depth sensing.
• Consequently, it is using neither a high definition map nor lidar to help the vehicle “see”.
• The issue as I see it is that while Tesla has access to far more data than everyone else put together, there is no evidence that it has solved the machine vision problem.
• In fact, the data that RFM has dug up indicates quite the reverse.
• A study by MIT (see here) provided data which puts Tesla towards the bottom of the pack.
• In this study, Tesla’s drove 112,427 miles in autopilot mode and suffered 18,800 disengagements or about 1 every 6 miles which is 1,846x more frequent than Waymo.
• This puts Tesla way below the leaders in 20th position which, ironically, is not that far from where it was when it last published data in 2017.
• In the last 6 months, most autonomous driving players have accepted that the problem is more complex than originally anticipated and have brought expectations of commerciality back to “some time in the next 10 years”.
• By contrast, Tesla thinks that its vehicles will be feature complete in H2 2019 and that they will be able to drive with the occupant no longer needing to pay attention by the middle of 2020.
• This is why Musk is at odds with his engineers.
• He seems to believe that AI is all-powerful which combined with his well-publicised fears around a dystopia where machines take over the world, is leading him to think that AI should be able to solve the autonomous driving problem fairly quickly.
• However, his engineers are coming up against the limitations of what deep learning is capable of leading to the realisation that the machine vision problem is far more difficult than Musk thinks it is.
• This is what I suspect is leading to the internal friction and the loss of so many key personnel from this division.
• While these problems persist, I think that it is very likely that those using Lidar and an HD Map will get there first because these tools greatly aid the system to properly understand the world around it.
• When machine vision is solved, I agree with Musk that lidar and the HD Map may become obsolete, but we are very very far from this becoming a reality as Tesla’s data indicates.
• Hence, the best-case scenario is that Tesla will get there around the same time as everyone else but given it is not using any of the tools that make the job easier, it may be late.
• I still think that while Tesla started the race to electric vehicles, it is unlikely to finish it.
• This is because it seems incapable of generating any cash from the expensive vehicles it does sell, meaning that when the OEMs get up to speed with their electric vehicles, it is likely to be crushed.
• I think Tesla will end up being acquired by one of the big OEMs who will then use it as their own electric vehicle offering.
• The price paid could well be less than the company is valued at today and so I would still steer clear.