Self-driving cars has long been one of the most exciting potential outcomes of advanced artificial intelligence. Contrary to popular belief, humans are actually very good drivers, but even so, well over a million people die on the roads each year. Globally, for people between 12 and 24 years old, road accidents are the most common form of death.

Google started its self-driving car project in January 2009, and spun out a separate company, Waymo, in 2016. Expectations were high. Many people shared hopes that within a few years, humans would no longer need to drive. Some of us also thought that the arrival of self-driving cars would be the signal to everyone else that AI was our most powerful technology, and would get people thinking about the technological singularity. They would in other words be the “canary in the coal mine”.

The problem of self-driving turned out to be much harder, and insofar as most people think about self-driving cars today at all, they probably think of them as a technology that was over-hyped and failed. And it turned out that chatbots – and in particular GPT-4 - would be the canary in the coal mine instead.

But as so often happens, the hype was not wrong – it was just the timing that was wrong. Waymo and Cruise (part of GM) now operate paid-for taxi services in San Francisco and Phoenix, and they are demonstrably safer than humans. Chinese companies are also pioneering the technology.

One man who knows much more about this than most is our guest today, Timothy Lee, a journalist who writes the newsletter "Understanding AI". He was previously a journalist at Ars Technica and the Washington Post, and he has a masters degree in Computer Science. In recent weeks, Timothy has published some carefully researched and insightful articles about the state of the art in self-driving cars.

Selected follow-ups:
https://www.UnderstandingAI.org/

Topics addressed in this episode include:
*) The two main market segments for self-driving cars
*) Constraints adopted by Waymo and Cruise which allowed them to make progress
*) Options for upgrading the hardware in a self-driven vehicle
*) Some local opposition to self-driving cars in San Francisco
*) A safety policy: when uncertain, stop, and phone home for advice
*) Support from the State of California - and from other US States
*) Comparing accident statistics: human drivers versus self-driving
*) Why self-driving cars don't require AGI (Artificial General Intelligence)
*) Reasons why self-driving cars cannot be remotely tele-operated
*) Prospects for self-driven freight transport running on highways
*) The company Nuro that delivers pizza and other items by self-driven robots
*) Another self-driving robot company: Starship ("your local community helpers")
*) The Israeli company Mobileye - acquired by Intel in 2017
*) Friction faced by Chinese self-driving companies in the US and elsewhere
*) Different possibilities for the speed at which self-driving solutions will scale up
*) Potential social implications of wider adoption of self-driving solutions
*) Consequences of fatal accidents
*) Dangerous behaviour from safety drivers
*) The special case of Tesla FSD (assisted "Full Self-Driving") and Elon Musk
*) The future of recreational driving
*) An invitation to European technologists

Music: Spike Protein, by Koi Discovery, available under CC0 1.0 Public Domain Declaration

Digital Disruption with Geoff Nielson
Discover how technology is reshaping our lives and livelihoods.

Listen on: Apple Podcasts   Spotify