[AnthonyMouse]

The obvious problem with steer-by-wire is that in the traditional design, it's not uncommon to lose power assist but not the mechanical connection to the wheels, so you can still steer the car. To completely lose steering control you'd need significant mechanical damage.

If the whole thing goes through the computer then there are lots of new ways to fail. Steering wheel position sensor goes bad on the highway? Computer gets bad data. Control wires get disconnected or damaged? No data. Completely unrelated wires get shorted and fry the computer? No steering. Anything pops the wrong fuse? No power, no computer or steering motors.

Some of those can be mitigated with redundancy but you're still vulnerable to common causes. You have three position sensors and someone dumps their beverage down the steering column, are there any left and do you have any good way to determine which one(s)? The vehicle took some minor damage allowing water to get somewhere it's not intended to, any way to guarantee you're not about to lose both sides of a redundant electrical system the next time it goes through a puddle infused with conductive road salt?

[spwa4]

Of course, a counterpoint is what's been happening in aviation. Autopilot became a thing. Autoland became a thing. And, to keep improving planes (first military planes, then commercial aircraft), it was much easier to drop the mechanical connection to the wings.

Autopilot started as a help to pilots, and evolved to something that is a necessity and pilot control inputs are "suggestions" or "goals", not inputs like turning the wheel on a bike. To be followed in what you might refer to as "the long term" from the perspective of controlling the aircraft, but in the short term, the computer is to fly the plane in a way IT thinks is reasonable. An extreme example would be to enforce the flight envelope. But today there exist autoland-only airports (as well as huge airports that go autoland-only if things are too hard for humans, like LHR)

Most of today's passenger aircraft cannot be flown if fly-by-wire is not operational. Most of today's aircraft actually used for passenger transport cannot land without fly-by-wire.

A number of military aircraft, and rocket planes and rockets, even the ones carrying humans, and more and more passenger planes cannot be flown by humans, not just because the mechanical force humans can generate cannot move the control surfaces (which "can be fixed" with hydraulics, if you don't mind serious caveats), but because the human brain is incapable of generating sufficient control inputs at a fast enough rate, or just can't keep stable flight going.

Hilariously, this also goes for hobby quadcopters. They are flown by algorithms. Humans can't do it. Not fast enough. Humans provide direction. Algorithms, even AI algorithms that aren't even guaranteed to succeed at all (in professional/military drones), actually fly the thing.

But, yes, you're entirely correct by saying "then there are lots of new ways to fail". It also works better, cheaper, faster, safer, more comfortable, ... if it doesn't fail.

And ... robotaxis are already far safer than even a good human driver. So whatever the problems ... they don't actually make things worse.

Also you should check out geohot's business. A lot of cars already are "fly-by-wire". Their solution? They now have 2 CAN buses instead of one. One for the critical stuff. Cylinder timings. Checking the oil levels. Turning the wheels. Actuating the brakes. That sort of stuff. A second CAN bus for your bluetooth music, and displays and what have you. I hear a certain new Mercedes now has like 7 buses. We are making things safer.

We can make this work. We will make this work.

[Scholmo]

Planes are probably the most controled machines we have. Everything gets checked twice or more, everything gets tracked and there is a clear requirement to do it like this because, as you said, its not possible for humans to control a fighterjet or a big plane.

Cars are non of that and we have billions of them on the street.

Cars also became a lot more expensive due to their complexity which def creates problems for a lot of people who can't afford all of that. I'm really torn by this because I think its very good that my side mirror shows me if there is a car next to me but in our capitalistic economy, we are excluding a lot of people from affordable cars. Drive by wire needs to be cheaper and easier to fix/repair.

Btw. Waymos are slowly learning to drive on highways so I might agree that they drive saver than humans in certain controlled envs. For sure not in any environment.

[spwa4]

But that is the "tradeoff" people are going for. What irritates me about Waymos is that they are not really cheaper than taxis and Uber. If we want people to become more mobile ... Waymo does not appear to be the answer.

And that was always the trade that was proposed. Sure, Waymo's (and Uber) will displace a LOT of taxi jobs, but they'll be way cheaper than taxis. Well ... they're not. And at that point, from an economic perspective, this is just taking things away for not much in return.

Once again people get a lot of possible choices and once again they choose for the more expensive one, putting more people out of business, out of a job, and as you say out of society. Now they're saying "yeah but this is good for autistic people and women, who can now travel by taxi without ever seeing anyone". How, exactly, does anyone think that's a good thing for society? Seriously?

Plus I'm a bit of the opinion, if Waymo is already breaking their own proposed social contract now ... imagine what they'll do in 10 years.