In cities at least I'm hoping that this also means that we no longer have half the width of every street taken up by parked cars. Instead they can all sit in (underground?) multi-storey car parks and be summoned at need.
There should be far fewer of them needed, too - at any given time what proportion of the cars in the UK are actually being used? And having them managed as fleets should unlock economies of scale as we all stop maintaining our own vehicles.
Driverless cars still need to be parked somewhere when they not in use. It would be hugely wasteful of fuel (or electricity) and result in congestion and extra wear and tear on the vehicle reducing its life span to just have them tooling around constantly when empty.
Rush hour will still be rush hour and that means everyone* commuting to work will need a car to get them there, so lots of cars on the road at the same time.
*Minus a few people who take public transportation or bike-- but that's true today too, and in the US very much a minority choice and likely to remain so.
I’m as enthused as you are Ed, about the future of driverless, however a several points of clarification and observation on your piece:
1) Uber and Lyft each have logged billions of cumulative rides, whereas Waymo has about 20 million. Waymo has surpassed Lyft in annual rides in San Francisco but to date the ride sharing companies overall have market share that is many orders of magnitude greater than driverless.
2) Waymo operates still in a sort of ‘trial’ mode where thousands of live human operators are monitoring rides from remote operating centers. Remote operations represent a drag on operating margins and as long as they persist, Waymo will be unprofitable and therefore cannot scale beyond select pilot markets.
3) from the standpoint of the user, to a large degree uber & Lyft already offer a ‘self driving’ experience. Absent Waymo’s soothing music and the Jaguar upholstery (BTW Waymo must surely be keeping Jaguar alive?) I can do anything in the back seat of an uber that I can with Waymo. Well…
A major difference is the convenience of very long rides like you envision on your vacation to France but even then I predict algorithms will prevent you from taking a journey to a destination where the probability of find a return customer is low…
4) finally, it is worth mentioning that major differences exist between the technology approaches of Waymo and Tesla. If Tesla’s vast library of roadway video interpreted by AI proves a viable substitute for Waymo’s hardware intensive LIDAR system, driverless will take off (and Waymo will fail) If not, ubiquitous driverless is likely to remain a long way off..
Two other points - I have a residence in Atlanta which is a Waymo test market. They can't drive on the freeways, apparently because the algorithms don't work well in high speed, high traffic environments, so you can't take one to the airport. Also, as someone who has used map applications for 20 years, in many different situations and countries, they often don't work well and give bad directions or suggest travel that no local person would recommend. That's puzzling, because if AI is so great, why can't it master map functions after 20+ years of effort? Map directions is one of the most perfect AI test cases you could possibly design and yet Google Maps frequently wants to route me down a "goat path" instead of taking the freeway in order to save me 15 seconds of travel.
I am not so sure. AI systems are/will be particularly effective at solving "closed system" problems, like a chess game with fixed boundaries, fixed numbers of pieces, exact rules for piece movement. Open systems, with no fixed boundaries, no fixed rules, unknowns entering in from the external world, are not amenable to algorithmic solutions. Freeway driving would be easy peasy if everyone drove the speed limit, used their blinkers when changing lanes, etc. But they don't and it's hare to write an algorithm for a system where people or processes don't follow the rules and deviate in unpredictable ways. The human brain is adapted to manage these type of incomplete information problems over 100 million years. I am not sure we can design AI systems to solve them anytime soon, if ever.
My source was a conversation with a Tesla investor several years ago but you are right to question that number. I just did a Gemini query and while Waymo’s early monitoring force was nearly 1:1, they have improved such that a single monitor can oversee ~150 cars So with today’s fleet of ~2500 cars, they are estimated to have 150-200 monitors. I was wrong to extrapolate their early coverage ratios to today’s fleet; they are certainly moving in the right direction. My personal experience with Waymo is three rides in Phoenix last spring, but as a VC investor in robotics I have kept an eye on the evolving driverless sector since inception. Gemini , Perplexity or Grok will serve up the primary sources I just consulted
There does need to be monitoring in the event of mechanical failure. Otherwise you have a car sitting in the middle of a road blocking traffic and no one the wiser other than other people being delayed by it.
The endpoint is using AI to assess real time analysis of “edge case” events before escalating to a human. We will get there but it requires a huge amount of compute.
We're not there yet., Witness AI moderation of social media. Completely innocent comments are flagged because they use a key word used in very different (and non-woke, non-offensive) context than the programmers thought of.
I move between San Francisco, London and the Cotswolds, and IMO rented autonomous vehicles are inevitable in dense cities, and just not ever happening in the exurbs, the countryside and long distance. Waymo's (optical + Lidar + Radar) ring-fenced urban model is more likely to succeed than Tesla's (optical-only) AI-everywhere approach.
I find Uber/Lyft too cringey, like I am intruding in someone else's private space, but love Waymos (and maybe soon Zooxs) for their anonymity.
Good reference to Le Shuttle there at the end... it always seemed like a slightly odd business model to be piling cars onto a train... but in a more futuristic setting it does feel a lot more like the ideal setup.
Fast, efficient train... cocooned personal transport pod with no weirdos present (apart from my own family).
Monstrous. The deep state will take my human driven diesel guzzler from my cold dead hands. Plus, it’s the only time the self-employed cannot feel guilty not working. I love driving.
Nothing says safeguarding like allowing your drunk daughter to pack herself into a taxi in 2025. On that basis, robotaxis cannot be rolled out quickly enough.
How much will congestion be reduced when we're all in driverless cars? Less lanes blocked by accidents, better use of the space on the roads. I imagine we'll all get around quicker.
Congestion will be far worse because automated vehicles will drive in an ultra cautious manner. People hugely underestimate how good humans are at driving at a collective level.
Re: Very soon, you will see fleets of driverless cabs making their way across busy intersections, as is common in many American cities.
I have never yet seen a single driverless car let alone "fleets" of them. And I live in a pretty major metro area (Tampa Bay). Earlier this year I rode in a semi-autonomous car a friend has. It was unnerving to say the least. Thank you, but I vastly prefer being the one driving the car.
Jon, we are in a minority here but the positives of driverless cars seem sparse unless you are actively anti car which I’m not. If taken too far, and it will be, humans will not be allowed to drive at all, eventually, or indeed do anything a machine can do more safely ) with cars (which won’t be owned ) being limited along eco lines ie far less autonomy and freedom and far more pointless. pass ag state control.
In terms of the "AI dystopias" that I grew up with in the 2000s, the world may be heading to something less along the lines of The Matrix or The Terminator series and more of Wall-E or Idiocracy. Machines in the future will do what their role has always been, making things faster and easier for lazy people!
I really don't want people mown down by badly driven motor cars but I enjoy driving and this seems yet another development that will rob us of the fun of doing things ourselves. I recently got a new car which had to be an automatic as they have practically stopped selling manual transmission ones (which in itself removes a good part of the pleasure). And it beeps at me constantly. It is fantastically annoying. So admittedly I might have difficulty meeting the eye of a grieving relative but I hope to keep doing my own driving for as long as I can.
If the trajectory of driverless cars is as promising as this essay suggests, then how long before 'safetyist' politicians and NGOs are proposing laws to ban humans from driving entirely?
That’s the trajectory! Cars were great liberators in the olden days and the olden days were all bad and so cars are suspect and must be restricted and eventually removed
I'm not worried about that in the US. People won't go for it. Consider all the fuss and furor we have over the notion that maybe we should require only electric stoves and ovens in new construction. Or even the fact that the US public remains dead set against joining the rest of the planet on the metric system
The 90% reduction is based on a new technology just being tested, the reduction should be much higher when it is a mature technology and they are dealing with fewer human drivers.
In cities at least I'm hoping that this also means that we no longer have half the width of every street taken up by parked cars. Instead they can all sit in (underground?) multi-storey car parks and be summoned at need.
There should be far fewer of them needed, too - at any given time what proportion of the cars in the UK are actually being used? And having them managed as fleets should unlock economies of scale as we all stop maintaining our own vehicles.
rows of parked cars also hugely decrease the aesthetic quality of streets
Driverless cars still need to be parked somewhere when they not in use. It would be hugely wasteful of fuel (or electricity) and result in congestion and extra wear and tear on the vehicle reducing its life span to just have them tooling around constantly when empty.
A system of small urban warehouses, multi-level, nearby to busiest parts of town?
Need far fewer cars
Rush hour will still be rush hour and that means everyone* commuting to work will need a car to get them there, so lots of cars on the road at the same time.
*Minus a few people who take public transportation or bike-- but that's true today too, and in the US very much a minority choice and likely to remain so.
You use your car no more then 10% of the time. I use mine on average 20 min/day. Huge waste of capital resources.
Great point. In big cities especially.
I’m as enthused as you are Ed, about the future of driverless, however a several points of clarification and observation on your piece:
1) Uber and Lyft each have logged billions of cumulative rides, whereas Waymo has about 20 million. Waymo has surpassed Lyft in annual rides in San Francisco but to date the ride sharing companies overall have market share that is many orders of magnitude greater than driverless.
2) Waymo operates still in a sort of ‘trial’ mode where thousands of live human operators are monitoring rides from remote operating centers. Remote operations represent a drag on operating margins and as long as they persist, Waymo will be unprofitable and therefore cannot scale beyond select pilot markets.
3) from the standpoint of the user, to a large degree uber & Lyft already offer a ‘self driving’ experience. Absent Waymo’s soothing music and the Jaguar upholstery (BTW Waymo must surely be keeping Jaguar alive?) I can do anything in the back seat of an uber that I can with Waymo. Well…
A major difference is the convenience of very long rides like you envision on your vacation to France but even then I predict algorithms will prevent you from taking a journey to a destination where the probability of find a return customer is low…
4) finally, it is worth mentioning that major differences exist between the technology approaches of Waymo and Tesla. If Tesla’s vast library of roadway video interpreted by AI proves a viable substitute for Waymo’s hardware intensive LIDAR system, driverless will take off (and Waymo will fail) If not, ubiquitous driverless is likely to remain a long way off..
Two other points - I have a residence in Atlanta which is a Waymo test market. They can't drive on the freeways, apparently because the algorithms don't work well in high speed, high traffic environments, so you can't take one to the airport. Also, as someone who has used map applications for 20 years, in many different situations and countries, they often don't work well and give bad directions or suggest travel that no local person would recommend. That's puzzling, because if AI is so great, why can't it master map functions after 20+ years of effort? Map directions is one of the most perfect AI test cases you could possibly design and yet Google Maps frequently wants to route me down a "goat path" instead of taking the freeway in order to save me 15 seconds of travel.
presumably the high speed/traffic problem will be overcome at some point
I am not so sure. AI systems are/will be particularly effective at solving "closed system" problems, like a chess game with fixed boundaries, fixed numbers of pieces, exact rules for piece movement. Open systems, with no fixed boundaries, no fixed rules, unknowns entering in from the external world, are not amenable to algorithmic solutions. Freeway driving would be easy peasy if everyone drove the speed limit, used their blinkers when changing lanes, etc. But they don't and it's hare to write an algorithm for a system where people or processes don't follow the rules and deviate in unpredictable ways. The human brain is adapted to manage these type of incomplete information problems over 100 million years. I am not sure we can design AI systems to solve them anytime soon, if ever.
Already solved in the Bay Area. The real challenge will be the chaotic traffic/human/animal streets of India.
AI is artificial but not intelligent
Ha, ha - my line is "only one of the words in AI are true"
Waymos are currently on Bay Area freeways.
Please cite your sources for "thousands of live human operators are monitoring [Waymo] rides from remote operating centers".
It doesn't sound like you have ever actually ridden in a Waymo - have you?
My source was a conversation with a Tesla investor several years ago but you are right to question that number. I just did a Gemini query and while Waymo’s early monitoring force was nearly 1:1, they have improved such that a single monitor can oversee ~150 cars So with today’s fleet of ~2500 cars, they are estimated to have 150-200 monitors. I was wrong to extrapolate their early coverage ratios to today’s fleet; they are certainly moving in the right direction. My personal experience with Waymo is three rides in Phoenix last spring, but as a VC investor in robotics I have kept an eye on the evolving driverless sector since inception. Gemini , Perplexity or Grok will serve up the primary sources I just consulted
There does need to be monitoring in the event of mechanical failure. Otherwise you have a car sitting in the middle of a road blocking traffic and no one the wiser other than other people being delayed by it.
The endpoint is using AI to assess real time analysis of “edge case” events before escalating to a human. We will get there but it requires a huge amount of compute.
We're not there yet., Witness AI moderation of social media. Completely innocent comments are flagged because they use a key word used in very different (and non-woke, non-offensive) context than the programmers thought of.
I move between San Francisco, London and the Cotswolds, and IMO rented autonomous vehicles are inevitable in dense cities, and just not ever happening in the exurbs, the countryside and long distance. Waymo's (optical + Lidar + Radar) ring-fenced urban model is more likely to succeed than Tesla's (optical-only) AI-everywhere approach.
I find Uber/Lyft too cringey, like I am intruding in someone else's private space, but love Waymos (and maybe soon Zooxs) for their anonymity.
Good reference to Le Shuttle there at the end... it always seemed like a slightly odd business model to be piling cars onto a train... but in a more futuristic setting it does feel a lot more like the ideal setup.
Fast, efficient train... cocooned personal transport pod with no weirdos present (apart from my own family).
Now we're talking!
lol
I know I'm a dreamer on this front, but one day this should all be possible.
I love going to France but the drive is too much.
Monstrous. The deep state will take my human driven diesel guzzler from my cold dead hands. Plus, it’s the only time the self-employed cannot feel guilty not working. I love driving.
Agreed
What is this obsession with always 'going somewhere'? Just stay at home and watch the light move across your carpet.
Nothing says safeguarding like allowing your drunk daughter to pack herself into a taxi in 2025. On that basis, robotaxis cannot be rolled out quickly enough.
How much will congestion be reduced when we're all in driverless cars? Less lanes blocked by accidents, better use of the space on the roads. I imagine we'll all get around quicker.
yeah, the co-ordination problem is presumably fixed a bit.
There will still be traffic jams- that's just the physics of turbulence which applies to all flows no matter how directed.
Congestion will be far worse because automated vehicles will drive in an ultra cautious manner. People hugely underestimate how good humans are at driving at a collective level.
Physics! Also known as the impenetrability of physical objects.
And wait for the car not to go bc you have posted a mean tweet or you have exceeded your meat allowance for the month.
Re: Very soon, you will see fleets of driverless cabs making their way across busy intersections, as is common in many American cities.
I have never yet seen a single driverless car let alone "fleets" of them. And I live in a pretty major metro area (Tampa Bay). Earlier this year I rode in a semi-autonomous car a friend has. It was unnerving to say the least. Thank you, but I vastly prefer being the one driving the car.
Jon, we are in a minority here but the positives of driverless cars seem sparse unless you are actively anti car which I’m not. If taken too far, and it will be, humans will not be allowed to drive at all, eventually, or indeed do anything a machine can do more safely ) with cars (which won’t be owned ) being limited along eco lines ie far less autonomy and freedom and far more pointless. pass ag state control.
We are still "allowed" to ride bikes, ski and engage in a number of other risky activities.
And ride horses! Lots of people love riding horses, but I'm glad we don't need them as such.
I could see authoritarian regimes limiting car use, but I don't think it's a huge fear in the West.
In terms of the "AI dystopias" that I grew up with in the 2000s, the world may be heading to something less along the lines of The Matrix or The Terminator series and more of Wall-E or Idiocracy. Machines in the future will do what their role has always been, making things faster and easier for lazy people!
Wall-E seems pretty accurate, although didn't anticipate Ozempic!
I really don't want people mown down by badly driven motor cars but I enjoy driving and this seems yet another development that will rob us of the fun of doing things ourselves. I recently got a new car which had to be an automatic as they have practically stopped selling manual transmission ones (which in itself removes a good part of the pleasure). And it beeps at me constantly. It is fantastically annoying. So admittedly I might have difficulty meeting the eye of a grieving relative but I hope to keep doing my own driving for as long as I can.
If the trajectory of driverless cars is as promising as this essay suggests, then how long before 'safetyist' politicians and NGOs are proposing laws to ban humans from driving entirely?
That’s the trajectory! Cars were great liberators in the olden days and the olden days were all bad and so cars are suspect and must be restricted and eventually removed
I'm not worried about that in the US. People won't go for it. Consider all the fuss and furor we have over the notion that maybe we should require only electric stoves and ovens in new construction. Or even the fact that the US public remains dead set against joining the rest of the planet on the metric system
I always say, it's a great idea until someone trips over the cord. And if the power is cut off can the passenger unlock the door to get out?
The 90% reduction is based on a new technology just being tested, the reduction should be much higher when it is a mature technology and they are dealing with fewer human drivers.
Chuckled much at the last paragraph 😁