Driverless public transport?
-
I just found out my local bus company is starting to operate driverless busses - this has come faster than I expected.
I will be interested to see how the pilot goes.
I know that many places already use driverless trains, but there is a big difference between a dedicated rail and when it has to share the road with other vehicles.The cynic in me thinks that even though they will eventually not need to pay a driver, I doubt the tickets will cost any less.
Here's a link for anyone interested - https://www.cavforth.com
Would you want to use driverless public transport?
-
As with all new technology I'm sceptical by default, until it's proven itself to work.
There have been some pilots here with small passenger vehicles driving very very slowly in public spaces with only pedestrian traffic. These things are mostly just gimmicks - it would in most cases just be faster to walk. So far I've not heard about one hitting anyone though, so that's promising
I certainly will not be queuing up to be one of the first to jump into one of these "deathtraps"
-
Tesla Self-driving 11.3.6 from San Francisco to Corte Madera with zero input.
Driving in the rain at night on city streets.
Latest version is 11.4.1, rolling out to some employees. I fully expect to see a wide adoption this year, and the number of Tesla drivers using FSD will scale rapidly. 2024 should see it come to Europe.
Tesla FSD is not limited to certain routes. It can drive anywhere using vision and map data, utilising Neural Networks.
-
@Pathduck I feel a bit more optimistic about this trial route because it will be operating a decent distance on motorways with no pedestrians or cyclists to contend with.
In those circumstances, the motorways have crash barriers either side by default, and a heavy bus will have more inertia than all the other cars around it, so it's probably safer to be in the bus than another car on the road
-
-
@LonM, I don't see such a big problem in public transport with a fixed route , this is perfectly acceptable for an AI with relatively low speeds in urban transport, it's not the same as in a car with random routes and speed in unknown areas, where there can be confusion on long trips. There the requirements for an AI are much higher. Multiple accidents with Teslas show it and they themselves have already stated that their promotions have been made on pre-programmed and controlled routes, which naturally were cleared without incident.
-
@Catweazle Safety is enhanced when Autopilot is engaged. (Source: Tesla 2022 Impact Report.)
Number of Vehicular Accidents
Per Million Miles (2022)- 0.18 Autopilot Engaged (mostly highway miles)
- 0.31 FSD Beta Engaged (mostly non-highway miles)
- 0.68 No Active Safety Features Engaged
- 1.53 Total U.S. Vehicle Fleet (based on data of Tesla vehicles and NHSTA data of vehicles)
Many so-called accidents involving autopilot were driver error, e.g. pressing the accelerator instead of the brake, which will override autopilot, or driver failing to pay attention (as required when using autopilot).
-
@Pesala, anyway in urban públic transport with fixed routes and relativ low speed, self driving AIs are less problematic.
-
One should be very careful with these things.
For me, the question is whether this is desirable at all.
In any case, Autopilot is far from being as mature as Elon Musk keeps making it out to be.
Recent information proves this once again.
-
@ingolftopf Of course, self-driving cars should be very careful, and they are — much more so than human drivers.
Autonomous cars are already a reality. Tesla FSD vs Waymo.
The big difference between Tesla FSD and Waymo, is that Waymo is ring-fenced to certain areas, and cannot drive on highways. The general solution is much harder to solve, but the advantages are immense. FSD has now gone out to a couple of cars in Australia and Europe, and to all eligible cars in North America.
Recent misinformation proves nothing. It focuses on individual cases, not the full statistical set of billions of miles driven on Autopilot. FSD is a step-change beyond Autopilot.
-
In recent news, a Cruise robotic car runs over a pedestrian, but one should not miss the point that she was knocked into the robot-car’s path by a human driven car. The driver failed to stop. The best collision avoidance system could not always avoid a collision. The Cruise vehicle did the right thing and stopped a.s.a.p. Unfortunately, the pedestrian was then trapped under the vehicle.
-
@Pesala, accidents happen and are not avoidable, especially if there is human interference.
An advanced AI, by definition, makes much fewer errors than a human (technical failures aside).
But this is valid for a well-developed AI, which in the past was not always the case, introducing an autopilot too soon in cars, IMHO, with resulting accidents in a traffic of human drivers too unpredictable and chaotic, where there are no exist the ideal situations, provided for in the tests. -
Adding too soon AI in cars
https://www.washingtonpost.com/technology/interactive/2023/tesla-autopilot-crash-analysis/ -
I just watched a fascinating video that suggests that driverless vehicles may well take us in the wrong direction with our urban areas.
https://www.youtube.com/watch?v=040ejWnFkj0
It gets a bit dystopian in the middle, but I think the core thesis is a good one: Driverless cars are not the solution for good towns and cities, what we need are carless drivers.
-
-
@LonM I follow JWZ's blog, he lives in SF and Waymos and other companies' cars are causing havoc on the streets regularly.
And what happens when one of these cars hurt or kill a pedestrian, and the billion-dollar companies with their armies of lawyers shrug and say it wasn't their fault but the stupid human walking into the path of the murderbot?
"...giving billion-dollar companies tickets would've just become a cost of doing business."
- Murderbot manufacturers still immune from legal consequencesThe future desired by robocar people is that in a concert hall of 40,000 people, 40,000 private cars just slow-drive around the concert hall from the 7:30pm dropoff until the 11pm pickup time, blocking streets for a mile in every direction, so that they can be ready for their masters at a moment's notice. Think "world's biggest traffic jam, everywhere, all the time". The future!
- A murderbot stamping on a human face foreverAutonomous murderbot tries to drive through Chinese New Year street party, gets whole entire ass handed to it.
- Burn Robot Burn!Google would prefer that their competitors' cars kill more people than their own, because that makes LINE GO UP.
- Google says it is good for their business if their competitors' cars kill more people.The visitors don't just come at night. They come all day, right to the end of 15th Avenue, where there's nothing else to do but make some kind of multi-point turn and head out the way they came in. Not long after that car is gone, there will be another, which will make the same turn and leave, before another car shows up and does the exact same thing.
- How much more terrifying could the murderbots be? Waymo.The big problem is that these are billion-dollar companies and in the US such companies are basically exempt from any regulation or laws that control the behaviour of their vehicles. And they have the money + lawyers to endlessly lobby politicians who will obey because "the economy" and "jobs".
-
@Pathduck Hearing about these stories makes me very glad that until now these companies have not been as present / overt in the EU. But given the kinds of politicians we are seeing more of I fear that they may start to succumb to lobbying to make that kind of behaviour acceptable over here too.
I suppose a silver lining to the horrible stuff US folks have to put up with is at least a way of "airing the dirty laundry" for all to see, and provides evidence to argue against further expansion.
-
Good quote:
"Move fast and break things."
- Well, what if that broken thing will be your spine under a robotaxi? -
Waymo’s driverless cars are big, expensive, unintelligent, and geo-fenced. They will be phased out. Waymo will have to come up with smaller, cheaper vehicles, and better software to survive in the future market.
Tesla’s Cybercabs are tiny, seat two passengers with loads of luggage space for airport runs, have intelligent AI software, and can drive anywhere, including on highways and dirt roads. They should be introduced next year, and in mass production by 2026.
FSD is fundamentally different to Waymo’s system. It is trained on billions of miles of real-world driving, and interprets what it sees without the need for precision mapping and lidar.
In the future, Robovans will carry 20 passengers (14 seated) to solve the problem of transporting crowds to/from large venues; 40,000 people will need only 2,000 Robovans (or 1,000 if they can return within the hour for repeat trips). It is misinformation that 40,000 robotaxis will be needed.
Robotaxis are not exempt from traffic laws, but may be able to take advantage of bus lanes, multiple occupancy lanes, etc.