4 ms·

Having the same fatality statistic was not enough for Google to bring autonomous driving into production. Many people think that it should be at least an order

by xiphias 10y ago

Having the same fatality statistic was not enough for Google to bring autonomous driving into production. Many people think that it should be at least an order of magnitude safer than manual driving

semi-extrinsic 10y ago
Sure. The problem with this is statistics: you need to have your autonomous cars travel at least 100 billion miles before you can confidently say they're an order of magnitude safer. A normal car drives about 100 000 miles in 10 years. You then need one million autonomous cars driving around for a decade before you even know whether they're safer!
- jgalt212 10y ago
  That's very true, but if you know your product well enough, and you perform well conducted experiments, you can come up with a good estimate for this without actually having your cars travel 100 billion miles.
  - semi-extrinsic 10y ago
    Sure. The main issue with that is the unknown unknowns. In hypothetical cars an order of magnitude safer than today, we're talking about events so rare that a single (hypothetical) accident which kills all 7 passengers due to the Tesla misinterpreting an obscure Hungarian road sign meaning "bridge out ahead" will significantly influence your safety statistics.
    - xiphias 10y ago
      There are enough youtube videos with people falling asleep in Tesla on the highway and a Tesla switching lanes and almost causing a frontal crash. These are known problems. Maybe a billion miles are needed for a fatal crash, but for small mistakes it's much less.
- protomyth 10y ago
  > A normal car drives about 100 000 miles in 10 years. Can you give a source for that number as it seems a bit low given warranty numbers.
  - weaksauce 10y ago
    my insurance estimates the average driven per year is around 12k miles so 100000 miles in 10 years is about right or a little low.
- ringwalt 10y ago
  You don't need an equal sample size to determine if autonomous cars are safer. But if you're comparing to all the cars on the road today, you need quality data that matches the distribution of cars today. Google isn't going to prove that autonomous cars are safer if they only test in Palo Alto in ideal weather conditions.

by xiphias 10y ago

Having the same fatality statistic was not enough for Google to bring autonomous driving into production. Many people think that it should be at least an order of magnitude safer than manual driving

semi-extrinsic 10y ago

Sure. The problem with this is statistics: you need to have your autonomous cars travel at least 100 billion miles before you can confidently say they're an order of magnitude safer. A normal car drives about 100 000 miles in 10 years. You then need one million autonomous cars driving around for a decade before you even know whether they're safer!

jgalt212 10y ago
That's very true, but if you know your product well enough, and you perform well conducted experiments, you can come up with a good estimate for this without actually having your cars travel 100 billion miles.
- semi-extrinsic 10y ago
  Sure. The main issue with that is the unknown unknowns. In hypothetical cars an order of magnitude safer than today, we're talking about events so rare that a single (hypothetical) accident which kills all 7 passengers due to the Tesla misinterpreting an obscure Hungarian road sign meaning "bridge out ahead" will significantly influence your safety statistics.
  - xiphias 10y ago
    There are enough youtube videos with people falling asleep in Tesla on the highway and a Tesla switching lanes and almost causing a frontal crash. These are known problems. Maybe a billion miles are needed for a fatal crash, but for small mistakes it's much less.
protomyth 10y ago
> A normal car drives about 100 000 miles in 10 years. Can you give a source for that number as it seems a bit low given warranty numbers.
- weaksauce 10y ago
  my insurance estimates the average driven per year is around 12k miles so 100000 miles in 10 years is about right or a little low.
ringwalt 10y ago
You don't need an equal sample size to determine if autonomous cars are safer. But if you're comparing to all the cars on the road today, you need quality data that matches the distribution of cars today. Google isn't going to prove that autonomous cars are safer if they only test in Palo Alto in ideal weather conditions.