SoylentNews is people
SoylentNews
Losing even one in 10 customers would substantially reduce airlines' revenue. They don't make much money on each flight as it is; less income would likely cause them to shrink their service, flying fewer routes less frequently.
The problem wouldn't just be customers who chose not to fly. Some passengers might split trips between self-driving cars and airplanes, which would further reduce airlines' revenue. For instance, a person in Savannah, Georgia, who wants to go to London could choose to change planes in Atlanta—or take a self-driving car to the Atlanta airport, and skip the layover.
These changes could substantially change the aviation industry, with airlines ordering fewer airplanes from manufacturers, airports seeing fewer daily flights and lower revenue from parking lots, and even airport hotels hosting fewer guests. The future of driverless cars is appealing to consumers—which means the future of commercial flight is in danger.
A personal fondling session from a TSA agent named Brad, or 5 hours in your self-driving Mazda that your four-year old smeared peanut butter in?
(Score: 0) by Anonymous Coward on Wednesday June 12 2019, @03:57AM (2 children)
But they are no where near 6:1 vs average drivers, and don't believe anyone that tells you they are--look deeper for the people behind the "AI" who are still tweaking madly as edge cases surface. These are not the self-learning systems you have been led to believe. There has been a lot of hype out there, dig past it.
And it will be a good while before they can come close to a "good driver" -- one who doesn't engage in known risky behavior like substance abuse, distraction (phone/videos), over-tired, etc, and is probably something like 10x less likely to be in a serious accident than "average" drivers.
(Score: 2) by takyon on Wednesday June 12 2019, @09:49AM (1 child)
~6:1 is one claim, but you can also find claims of 90% or 93% accident reduction.
An individual car doesn't need to learn anything. It just needs to apply algorithms that have been widely tested. These algorithms may have been created using a "machine learning" process. If you don't want to call it "learning" to mix it up with some other method (like having neuromorphic "strong AI" enslaved in the car), then don't.
I don't think it will be long at all until it gets as good as a "good driver". >99% of the time, driving is trivial and boring. Google demoed a car doing the boring stuff years ago. But sometimes a person will run in front of the car or some other unexpected event will happen. In those cases, if the car can react very quickly in order to simply slow down, it will increase the possibility of saving lives (might result in you getting rear ended more often though). If it can react in 25ms instead of 500ms, there can be a substantial reduction in impact velocity.
https://www.science.org.au/curious/technology-future/physics-speeding-cars [science.org.au]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by JoeMerchant on Wednesday June 12 2019, @11:27AM
Whatever the odds (and never tell me), self-driving cars represent predictability, mass production with uniform predictable flaws and failure rates: the actuarial table calculators are dreaming of retiring to Bora Bora on this one, and the deep futurists can't stop salivating over the day that we get the human controlled vehicles off the road.
Newsflash: freeways might someday become auto-driver only zones, but until you're prepared to stop me from my morning bike ride, and keep pedestrians off the village crosswalks, you're always going to have human controlled obstacles in the auto-drivers' paths.
🌻🌻 [google.com]