I took a day off from posting on TMC to read Morgan Stanley's green paper on autonomous vehicles. It's good infotainment, but I think a good dose of skepticism is in order. I won't attempt a thorough critique, it is much too massive for that. But I will offer a few observations.
In general, I think MS is much too focused on a utopian state wherein all cars are autonomus, than in the more interesting questions of how we get there, where is the motivation coming from, and how big the pay offs will be at each intermediate state. They offer four phases. The first three are essentially advances in autopilot technology, to use Musk's distinction between autopilot and autonomous. These three phases should roll out over the next six years. The fourth phase, a utopia of autonomous vehicles will have to wait another 20 years. And yet the $1.3T payoff they envision only makes sense in this utopian state. They make no assessment of what benefit may accrue by 2020. This is problematic because the gains made by 2020 may be so great and the problems posed by forcing somebodies idea of a utopia may become visible enough that the utopia is laughed off in due course.
It is also curious that this paper takes such a dim view of EV prospects and substantially overlooks Tesla's potential in this space. They contemplate Google entering this space as Silicon Valley tech with no exposure to auto manufacturing, but ignore Tesla which even when this was written was clearly both Silicon Valley and an auto maker. Perhaps they gnored it because Tesla had not yet pushed any agenda toward autonomous vehicles, but that misses a very fundamental point about what Tesla is upto. Tesla is building a completely programmable car, a digital car, from the ground up. This is exactly the kind of transformation of the car that is needed on the path to full autonomy. Every aspect of the car must be within the awareness of and under the control of an integrated computing system. It's all about a platform for digital control. So reading the paper in this light reveals that Tesla has incredible potential to play multiple roles from hardware to softare, to interior experience integrator. The operating system that Tesla is building out could well lead to a service Tesla provides to other OEMs. If MS is correct that in the future the car value will be 40% hardware, 40% software, and 20% content, user experience, then what Tesla is developing on the software side may be just as valuable as say, the Gigafactory. This is astounding.
But I would argue that the software value and other other value that Tesla is creating along this path to auonomy does not require the utopian world they imagine. Consider what we have recently learned of Tesla's advances with autopilot and dual motors. In 12 months Tesla has leapfrogged over phase 1, well into phase 2. With lane change, summon and parking capabilities it is getting into phase 3. I suspect that they may actually be well into all phase 3 capabilities at this point, but simply need to to more reliability testing and refinement. I strongly suspect that they already have test vehicles that can navigate themselves to any preselected address. They are able to make so much progress within 12 months, not just because they've got great programmers and engineers, but because the programmers have a great platform to work with, a highly programmable architecture.
Consider this, the Model S already has a software layer between the driver and the mechnical systems of the car. The driver is only manipulating a controller, and software already transforms that input into action. This digital control affords much more that merely adding on safety features. Every aspect of performance can be modified through programming. So for example, dual motors, we now know, can be digitally controlled in such a way that 10 to 30 extra miles of range can be obtained. This involves having an algorithm to optimize the distribution of power from front to back. This algorithm could take in as inputs the current speed, the desired acceleration (from the driver), the angle of the steering wheel (also fro the driver), the curvature of thebroad ahead (from radar, camera and gps), the grade and slope of the road (precice maps and sensors), slippage in the wheels (from sensors), even weather conditions. All this can potentially be taken in to optimize power sent to fron and back motors. Now suppose programmers have just a simple model to work with. They program that up and tweak it until it works great. Future research perhapsnbased on real world data from customer carsnon the road may lead to formulating better models for integrating so many variable and deriving an optimal power assignment. No problem. Programmers gonto work on coding up the new model. It gets tested, and if it delivers better performance, it gets pushed out in a software upgrade. So now any Model S owners with sufficient hardware ge a performace upgrade over the air. This is much more than just delivering a certain safety feature or autopilot capability, this is about delivering high performance. Drivers will want Tesla's autopilot technology not because it lets them go to sleep at the wheel, but because it enhances their driving experience. Imagine being able to drive with the precision of highly skilled professional driver through a winding mountain road. You take each cure with confidence, you accelerate and brake in perfect coordination to make the most of every curve. You have incredible control the the whole way. The car responds impeccably to your slightes suggestion. All the while, it is the software that is intepreting your instruction, road and vehicle conditions and transforming it into something amazing. The potential of the digital car is not only to enhance the performance of the car, but the performance of the driver too. You will want to drive this car because it will feel amazing to be the driver. The P85D is amazing not because some professional driver can make it do 0-60 in 3.2 seconds, but because you can make it do 0-60 in 3.2 seconds.
So what I am talking about is what Tesla is delivering this this decade. The saftey gains from autopilot can be enormous. It shoul not depend on getting to some utopia to kick in. Two years ago a new model S driver fell asleep at the wheel, crossed over the opposite lane and struck a bicyclist. The bicyclist, a librarian at UCSC, died. This is exactly the sort of thing autopilot should be able to prevent now.
I've got more to say, but am running out of time. Next time.
