Hacker News Reader

Comment by Animats | Hacker News Reader

Cruise started as "fake it til you make it", with a drive down the Great Highway in San Francisco. The ideal public road - several miles of straight road with no cross streets. They managed to parlay that into an acquisition by GM and board seats on GM's board.

Then they didn't deliver.

Waymo hasn't left blood on the pavement. Cruise and Tesla have.

View on HN

Wildly misleading comment. Like none of this is true. Cruise started as a retrofit kit for audis similar to comma.ai then they pivoted to l4.

Here it is, the original Cruise RP-1 video.[1] There's the car, driving down the Great Highway in the Sunset district of San Francisco. Perfectly straight, no cross streets, great visibility.

[1] https://videos.dailymail.co.uk/video/mol/2016/03/11/70647585...

Nobody ever promised full autonomous with this kit to the best of my knowledge so your claim of "fake it till you make it" is a mischaracterization. It calls it "highway autopilot" in the video, shows highway, what is the problem?

Well, no human blood. A Waymo has hit a dog in what seemed like an unavoidable accident.

Waymo hasn't left blood on the pavement. Cruise and Tesla have.

As long as we're clutching our pearls and being overdramatic, how much blood have the humans left on the pavement in the meantime?

On a normalized basis in the USA? 1 fatality per ~4000-5000 years of driving. 1 injury per ~60-70 years of driving.

Can you point to scientifically rigorous evidence, at least reaching the bare minimum of fit for publication, of the absolute level of safety of those systems and where they stand in comparison?

How many people am I allowed to kill or injure before I need to run scientific studies on unproven and empirically fatal technology?

This rebuttal is fine, i guess, but to answer your final question: At least 1645 deaths if your government is british. I forget the US numbers, but globally you can kill upwards of 30,000 people with unproven and empirically fatal technology.

Really, you are arguing that I get to kill 1645 in the UK before I have to show any scientifically rigorous evidence? I get to do 1645 drives and kill the passenger every time before anybody cares?

That is not only insane, but empirically false in the USA as people cared very greatly about Cruise injuring people during the development process.

You have to be answering the question: “Assuming we already know for a fact that the technology is safer than a human driver, then, if scaled to all drives, how many can die annually and still be morally justifiable?”

My question is: “How many do I get to kill before I need to figure out if I am killing fewer people?” That is the basic question to be asked during the development process of any unproven dangerous technology to minimize sacrifices to technological dead ends.

i wasn't commenting on your specific example of unproven and fatal technology, just answering the general question with a bit of snark.

considering humans leave oceans of blood in their wake with their own driving, all of the autonomous driving companies are doing great by your measurement of success

There's always someone saying this, and they always don't seem to realize that random humans aren't a company with billions of dollars in investment. If your billion dollar investment kills someone and you are even possibly found liable, that can kill the whole company right there

Reply on news.ycombinator.com