Kevin Kelly on the Future of Autonomous Cars

Originally published at:

I still think they’re a bad idea. We dedicate a tremendous amount of space to cars already and have a strong pattern of adjusting our consumption of goods to increase with better convenience.


I believe autonomous cars will be here sooner than 15 or 20 years. The speaker is talking about cars that can operate perfectly. The cars don’t have to be perfect, they only have to be better than humans at driving. Humans suffer from all sorts of impairments. I believe that within 5 years that we will have a competent car. I also believe that around 10 years from now that insurance companies will only offer insurance to some people only if they have an autonomous vehicle.


I watched the video and was surprised how Mr Kelly really didn’t talk at all about the current state of autonomous driving technology, or explain in any technical detail why he thinks we’re going to need massive changes to public infrastructure before they can be possible. Someone watching this video without knowledge of any current testing might walk away thinking this was all purely a hypothetical discussion, not something that companies have already been working on for years.

As Tesla and Uber have demonstrated with their fatal crashes, most “self-driving” cars of today definitely aren’t ready for prime time. (Google’s Waymo seems to be closest but thankfully they’re taking a slow-and-steady approach). But to my knowledge it wasn’t signage or infrastructure issues that led to those fatalities. A car that’s got a good GPS, image recognition and access to a detailed LIDAR map can know exactly where it is and the appropriate speed limit without special signage added to roads. It’s the different “edge cases” such as extreme weather, unpredictable pedestrians, or weird unanticipated lighting conditions/software glitches etc. that seem to be the biggest hurdle at the moment. Maybe that’s wrong, but if we need all our roads rebuilt then Mr Kelly ought to spend a little more time on the technical reasons for that.


autonomous buses would be a neat convenience. and maybe we can put them underground so as not to disrupt pedestrians. and give everyone a card with credits on it to have access.

or we can do what Olympia, WA, is doing:

Scheduled bus service will be temporarily suspended and replaced by a reservation-based transit service to continue providing essential trips to the public. Dial-A-Lift services will continue to be available for qualified passengers making essential trips.

1 Like

They are already here. Maybe not fully autonomous, but it’s really splitting hairs. One of the flaws of predictive thinking is that you suppose certain conditions will need to be met for scenario b to follow scenario a (see: four laws of robotics). That’s not the case. Someone just has to sell it, whether it works perfectly or not. That’s it.

ETA: I actually have trouble envisioning a future in which cars are truly fully autonomous. Maybe as some sort of public infrastructure in dense urban environments, but if the vehicle has to deviate from compliant motorways at all, say into rural areas or private driveways and non municipal roads, there will need to be a takeover option for a very, very long time to come.

1 Like

Already many new cars have lane holding and following features that help to operate in rush hour traffic. My car has some of this and it is much less infuriating to assume more of a babysitting role rather than constantly accelerating or braking. Lane holding takes some of the monotony out of long, mostly straight stretches of highway. The current lane holding systems are dependent on clear, unbroken lane paint. I predict that laying some kind of wire in the roads will become as easy and common as painting lines. Full autonomy like getting in and saying “Take me to Costco” is quite a ways off as is any kind of operation in inclement weather.

1 Like

Funny you say that, as I was just reading accounts of people who are part of Waymo’s test program in Phoenix using the automated cars for Costco trips last year.

But yes, it’s still a test program and does currently have many limitations, including weather.


I hope that this technology is available to all of us before we get too old to drive safely.

1 Like

I don’t know about you, but I’m sure many people have already questioned my ability to drive safely since I got my license at 16. (No accidents yet though, fingers crossed)

1 Like

I think the problem is what you call edge cases happen continuously. One example: I was driving to work this morning & encountered a tree-trimming crew. Two trucks blocked the other side of the road. The person controlling traffic in my direction was a sleepy young man who was still trying to get his yellow safety vest on with one hand while yawning & holding a stop sign sort of tilted & at an angle in the other hand. I saw the whole scene, slowed to a stop & put on my flashers. Maybe I don’t understand how smart AI will be, but I’m skeptical a computer will ever be able to do what my brain did insantly to understand the situation & act accordingly. There are far too many ambiguities & variables. How would the computer understand what two large cherry picker vehicles carrying men with chainsaws meant? How would it know what the sleepy young man, near me but a hundred feet from the trucks, represented? It seems impossible.

So I think Mr Kelly implicitly is referring to the unending complexity of decision-making while driving when he says local roads need to be re-engineered to accept autonomous vehicle. Like interstate highways, they will have to be simplified & cleared of obstacles. Pedestrians will not be allowed. Protocols for unusual situations will have to be strictly followed. And it seems to me that autonomous vehicle roads will be only for autonomous vehicles, because mixing drivers & AI will be disastrous.


If we can barely write a bot that can produce an article summary, then self-driving cars must be a way off.

1 Like

I totally agree. A truly autonomous car needs a conceptual understanding of the world around it, something that all machine learning algorithms (please don’t call these things AI, that is just misleading) we have today are completely lacking.

I really hope that a future where pedestrians (and cyclists?) are forbidden on local roads will never become reality.


I briefly worked at an autonomous vehicle company as a software developer. After listening to Mr. Kelly, I don’t feel he is qualified to speak on this topic. To drag out that old barb from Wolfgang Pauli, Mr. Kelly is “not even wrong.”


I do like the assertion that the thoroughfare infrastructure will change to accommodate automated vehicles. We have railroads, bike lanes, and sidewalks. Maybe the innovation that should be happening is on personal self driving vehicles closer to bikes. If that doesn’t sound so great, what is so great about cars?

I’m not curious about the “If an autonomous car kills someone, who’s liable” problem or how to keep the AIs aligned with our human values; what I want to know about is the bigger questions of privacy and freedom. Having a car that you yourself drive and can take anywhere offers people a great deal of freedom and personal autonomy. That kind of freedom is very important for a whole lot of people, not just those who have cars they like to take on joyrides but for low-income people whose cars mean that they’re the only ones who are capable of driving their families to and from places, or their truck is necessary for them to do work, and so forth.

There’s also the issue with whether or not you’ll actually own your autonomous car. You’ll likely be renting it as a subscription from a company, and not be able to repair it or tweak it or mod it on your own, and it’ll be collecting data on where you go and what you do with the car’s wifi all the while.

So yeah, the future of autonomous/self-driving cars that I see is a phasing out of standard cars, with corporate lobbying campaigns pushing regular cars and drivers as “dangerous”, leading to even larger widening of the gap between rich and poor, with the rich having their self-driving pleasure pods and low-income people forced to ride bikes or e-scooters, or take underfunded and/or privatized buses and metros.

1 Like

That’s possible I guess, but one of the reasons that development of self-driving cars has more or less been put on hold by a lot of major car manufacturers is that the companies are realizing that there’s not necessarily a lot of profit to be made by switching their business model from “every family needs to buy one or more cars” to “people will use autonomous taxi services to get around and won’t buy their own cars anymore.” If there’s competitive autonomous driverless taxi services available that could actually end up costing less to use than owning and maintaining your own car, as well as less cars being built overall. So if that we’re to happen it’s not clear to me how that would widen the gap between rich and poor. Of course the rich will travel in their own private luxury pods, but that’s no different than the chauffeured vehicles or luxury cars or helicopter taxi services that they already travel in.

I still think the most transformative potential for self driving vehicles isn’t the added convenience it would bring to a majority of people, but the life-changing potential for people who currently have no ability to travel independently at all such as the blind or the elderly.

I have long maintained that the most logical place to start rolling out these technologies is in retirement communities. You have all the right circumstances: limited range in a highly controlled environment, private roads that require less government oversight, low speed limits and a high-demand population.

I too am bearish on Level 4 or 5 self-driving cars any time soon, perhaps even never.

Best evidence against is the recent self-inflicted dick punch Telsa gave themselves when they turned on “Summon Mode”. The task of automatically getting from a parking slot to a hotel front door failed lots. YouTube has evidence. Well, videos.

Not trying to be snarky, but… citation? Or subscription deets re: your newsletter.