🚗 Uber launches self-driving cars in a trial in Pittsburgh this week. Read Max Chafkin’s excellent exclusive on this and their acquisition of Otto,
the self-driving truck company. I was in an Uber returning from the airport when this was reported on the BBC by EV subscriber, Rory Cellan Jones. My Uber driver heard the headline and leant forward to increase the volume, ears pricking up. This is the sharp end of automation.
Even if you hold the reasonable belief that work isn’t going to go anywhere soon, and that we’ll continuously reinvent things for humans to do, the question is not about the statistics in aggregate. It’s about the (quite significant) numbers of individuals who will be affected by the transition.
What do soon-to-be-former Uber drivers do? And how far will their wages be suppressed as self-driving cars roll out? (See the robotisisation of work piece at the top of this week’s EV.)
Another lens for Uber’s announcement is how they have o'erleapt Google in getting a product to this stage of testing. Google was early with on-street self-driving cars but their project seems to have hit snarls with execs leaving
For Uber, mastering self-driving cars is an existential need. Whereas for Google, it is a nice-to-have.
Part of the challenge for Google is also about the pace of innovation in both hardware and software (particularly the GPU/algorithm stack powering new vision applications). Much of Google’s own self-driving choices would have predated innovation in that area, facilitating new entrants who can leapfrog.
💥 For a great illustration of this, see Nvidia’s deconstruction of end-to-end deep learning for self-driving cars
, where they demonstrate “that CNNs are able to learn the entire task of lane and road following without manual decomposition into road or lane marking detection, semantic abstraction, path planning, and control. A small amount of training data from less than a hundred hours of driving was sufficient to train the car to operate in diverse conditions, on highways, local and residential roads in sunny, cloudy, and rainy conditions. The CNN is able to learn meaningful road features from a very sparse training signal (steering alone).” EXCELLENT, MODERATELY TECHNICAL
Increasing numbers of self-driving EVs should put pressure on existing automotive business models. In particular, the sales and aftermarket and second-hand car business. For one thing, even first generation EVs are much for reliable than ICE vehicles. 🔮 Check out this report of the maintenance required on a Tesla Model S
which had driven more than 100,000 miles. (Diddly-squat is a fair approximation.) And, of course, shared ownership, transportation-as-a-service applies pressure to the whole ecosystem of local mechanics, second-hand car sales and parking.