Hacker News Reader

Comment by ayanray | Hacker News Reader

What I worry about is how can we define accountability. At least you could sue a person if you got into a car accident and got injured (happened to several people I know). Suing a huge corporation and getting bullied around, settling for less, etc. sounds possible, but can happen too with people vs people. How do you even make a case when you likely don't understand what actually happened or could even prove what happened (crypto, copyright laws)? I'm all for reducing risk, but machines will make mistakes and I don't know what happens next.

View on HN

History seems to have shown the opposite, in transport.

Trains are engineered to absurdly high safety standards and every time there is a crash, it's widely publicized and an in-depth investigation is required along with suggestions for systemic changes to prevent repeat events, and the companies are required to pay large settlements:

http://www.npr.org/sections/thetwo-way/2016/10/27/499592760/...

At the same time, trains are vastly safer than personal cars. Yet when a driver kills someone, police can barely be bothered to attend the scene and rush to excuse the driver and let traffic flow again.

The safety standards on vehicles are such that billions are spent on engine development yet we tell people to watch out for the "blind spot" in that trucks mirror contraption. We're soon going to have autonomous driving AI yet trucks are roaming city streets that are unsafe by design and regularly claiming lifes. I don't see Volvo and MAN settling with anyone over that.

At the same time, trains are vastly safer than personal cars.

Most likely because trains essentially only travel in one dimension almost all the time, along a very well-defined path.

That's the point. For their base (and empirical) level of risk, there are huge binders of safety regulations for trains. Yet cars, which pose a vastly higher risk, are basically unregulated and safety is commonly ignored for what are peripheral concerns.

I worry this is a case of micro vs macro environments. In a macro environment, it would be of interest to the general public to not determine why 1000 people died in a train crash. It's worth an exhaustive multi-year study and the victims' families can join a class action lawsuit and be tried as a single case. It's also in the best interest for all parties to find out why that crash happened in such a controlled environment to improve the service (private use of railways, well-understood environments).

In car crashes, each case could be different and difficult to figure out what was the exact cause. Cars are becoming more complicated, as with farm equipment vs farmers or the "check engine light". Without transparency into the internals, I wouldn't know what broke or was it my fault or the car and I would probably have to rely on the honesty of the company to tell me or my family that.

I think in order for you to 'sue' someone - they have to be negligent.

For the same reason generally passengers don't sue airline companies when something goes wrong.

There is an unbelievable amount of energy and money that goes into safety ... so much so that flying would probably be 1/4 the cost if they simply didn't care that much about it, and so there's a degree of 'public goodwill' in the whole thing.

I think the 'moral' issues will come down to scenarios wherein the computer has to make a decision like: do I swerve and kill my 2 passengers, or hit the car in front of me head-on and kill it's 2 passengers: pick one of two outcomes. I think this is the moral dilemma that is already coming up.

Reply on news.ycombinator.com