ResearchGate Twitter Flickr Tumblr

Bill Gates Again Manages to Keep AI in Perspective (And Autonomous Driving)

Earlier this week, I wrote about how Bill Gates isn’t carried away by AGI nonsense, while “Microsoft Research” simultaneously went off the rails at with a publication that can charitably be called a badly written sf novelette disguised as a paper.

I still had GatesNotes open in a browser tab, and last night a new entry popped up on “The Rules of the Road Are About to Change,” with a take on AI for autonomous vehicles that is interesting for two different reasons.

First, there’s the LLM-related approach to training:

When you get behind the wheel of a car, you rely on the knowledge you’ve accumulated from every other drive you’ve ever taken. That’s why you know what to do at a stop sign, even if you’ve never seen that particular sign on that specific road before. Wayve uses deep learning techniques to do the same thing. The algorithm learns by example. It applies lessons acquired from lots of real world driving and simulations to interpret its surroundings and respond in real time.

The result was a memorable ride. The car drove us around downtown London, which is one of the most challenging driving environments imaginable, and it was a bit surreal to be in the car as it dodged all the traffic. (Since the car is still in development, we had a safety driver in the car just in case, and she assumed control several times.)

I think this is a nifty approach; I can imagine it will push things forward quite a bit.

Then, Gates’s personal predictions on autonomous driving are equally interesting, not least because it’s refreshingly free of AI hype:

Right now, we’re close to the tipping point—between levels 2 and 3—when cars are becoming available that allow the driver to take their hands off the wheel and let the system drive in certain circumstances. The first level 3 car was recently approved for use in the United States, although only in very specific conditions: Autonomous mode is permitted if you’re going under 40 mph on a highway in Nevada on a sunny day.

At Level 3 (SAE 3), to recall, the driver can have their “eyes off” the road and busy themselves with other tasks, but must “still be prepared to intervene within some limited time.” The level of automation can be thought of “as a co-driver or co-pilot that’s ready to alert the driver in an orderly fashion when swapping their turn to drive.”

Make no mistake—this step from SAE 2 to SAE 3 is a monumental one. However, just like other hypey AI hypes, SAE 3 too isn’t right around the corner. Gates again:

Over the next decade, we’ll start to see more vehicles crossing this threshold. […]

A lot of highways have high-occupancy lanes to encourage carpooling—will we one day have “autonomous vehicles only” lanes? Will AVs eventually become so popular that you have to use the “human drivers only” lane if you want to be behind the wheel?

That type of shift is likely decades away, if it happens at all.

If SAE 3 really happens “over the next decade,” that would be in the ballpark of what I consistently (and insistently) predicted around seven or eight years ago—that even SAE 3 as autonomous “co-pilot” driving would take fifteen years of development and infrastructural changes at least to become feasible. (A prediction that got screamed at a lot at the time, metaphorically speaking.)

But all that notwithstanding, I do not think that the scenario of autonomous individual cars will serve us well. In 2019, Space Karen took a swipe at Singapore’s government for not being “welcoming” to Tesla and not “supportive of electric vehicles,” which is complete bullshit in and of itself, of course. (I visited CREATE in 2019, Singapore NTU’s Campus for Research Excellence and Technological Enterprise, and the halls were filled to the brim with research projects on autonomous electric transportation.) But in Singapore, the focus is on public and semi-public transportation, not private cars. And I gotta say, I loved it when the Singaporean secretary for the environment and water resources, Masagos Zulkifli, shot back by saying that Singapore is prioritizing public transportation: “What Elon Musk wants to produce is a lifestyle. We are not interested in a lifestyle. We are interested in proper solutions that will address climate problems.”


Thus, while I’m pretty excited about advances in autonomous driving, I’d be even more excited about advances in autonomous driving toward sustainable transportation.