clacke
@PhilSalkie @azonenberg I used to work at an automobile safety company where some of the products were things like automatic lane-keeping. I didn't work directly on the products, I was there to fix their build system.
They were working on full autonomy, and in theory had it, but were working under the assumption that nobody will allow fully autonomous cars on the roads until they are 10× safer than human drivers, and that wouldn't be happening until 2030 at the earliest.
clacke
They were struggling with things like "we have to pass this certification, but if we optimize for stopping for the fake child, we will be worse at stopping for a real-world child". They were taking the ethical issues seriously.
Then Musk suddenly went and released fully autonomous driving while pretending that's not what he was doing, bypassing legal issues and safety issues and putting all blame on the human driver. It was *worse* than what other companies had, and from what I gather it is *worse* than human drivers, not better, and certainly not 10× better.
At this point I am not convinced that autonomous cars will ever be safer than humans. Certainly not autonomous cars built by commercial companies competing in a time-to-market race to the bottom. Anyone taking safety seriously will be out-competed.
It's just move fast and break things plus deadly traffic incidents.
clacke
To make truly safe and effective autonomous cars, I believe the traffic environment has to change, to accomodate the cars. Special markings, signage, separation, etc. In the end, human drivers will have to be banned.
At some point in this transition, we might as well have gone for more efficient rail-bound PRT instead of trying to fit automated traffic into our made-for-humans, mixed-with-humans traffic environment.
Andrew Zonenberg
@clacke @PhilSalkie Autonomous vehicles in closed environments like rails or tunnels are completely plausible, but I don't see them ever being viable in the general case especially in the kind of lower density / rural areas I spend a lot of my time in.
Last time we had a major snow storm, the public works department ran out of road-closed signs to block off all of the steep hills they lacked the manpower to plow. There's zero chance of them ever getting the resources needed to make and maintain robot-readable markings everywhere.
What we're going to get in the meantime is half-baked solutions made by the same people who think LLMs are going to replace human writers and scientists and who knows what else.
And who are running it all on hardware built by the lowest bidder without any effort put into reliability.
It is indeed "move fast and break things". The cars are moving fast and they're breaking things.
Linux Walt (@lnxw37j1) {3EB165E0-5BB1-45D2-9E7D-93B31821F864}
clacke
@forthy42 I agree, but currently it's not even better, but people still use it. Because they can.
Bernd Paysan R.I.P Natenom 🕯️
@azonenberg @clacke @PhilSalkie A lot of data for roads and autonomous cars should be in a public accessible GIS. What speed is where, where are road construction sites, what has been plowed and what not (updated in realtime by the snowplow and the weather radar, which knows how much snow is coming down).
Trying to drive an autonomous car like a human does is silly.
However, since humans driving kill thousands each year, self driving has a low bar. If 10 times better, it's good enough.
clacke
Bernd Paysan R.I.P Natenom 🕯️
@clacke @azonenberg @PhilSalkie One of the dangers of computerizing anything is that people overestimate the capability. The computer told us, so it must be correct.
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.