On Mon, Apr 6, 2015 at 6:18 PM, Lee Hart via EV <ev@lists.evdl.org> wrote:
> Here's the "elephant in the room" that I haven't seen addressed: When a > self-driving car is involved in a fatal accident, who pays? Who goes to > jail? > > I wouldn't want to be in a self-driving car because when (not "if") > something goes wrong, every personal injury lawyer in the country will be > filing lawsuits, against anyone and everyone even remotely involved. Lee's quite right of course. Everyone is speculating on what this is now, what it will evolve into, and generally thinking in broad strokes. But it's the details that will bite hard. Simple sense-and-response control would be just begging for a tragic outcome. And there is no AI system sufficiently advanced to make the right decision in every case. There will be wrong responses, some tragic. Also, if you cede too much control to "the system" that means you have little control when some glitch becomes a seriously FUBAR situation. Making this work really well would require a massive software validation effort that few companies will do properly. (They frequently have a hard time implementing CANBUS properly.) This is all a perfect example of the "four wheel drive" analogy - four wheel drive can allow you to do things that are otherwise impossible. But if it is implemented or used improperly, it will only get you deeper into trouble than you might otherwise have been in. Sadly, there will be lawyers that specialize in these cases, and they will make lots of money. There's a far stronger case to be made for using tech in ways where we know it works - sensing what is difficult for humans to perceive, and rapid response. I would welcome an infra-red HUD that would allow me to see in the dark or through fog, dust and blizzards. Or an audible warning that closure rate to the vehicle ahead is too fast. Maybe even automatic braking in that case, but I'd have to try it first. An audible warning that you're about to collide with someone in your blind spot would be good, but a steering correction would NOT be OK. Like if I'm purposely changing lanes into someone because that collision is the lesser of two evils. That's a decision I want to make myself. I'm all for enhancements. But the decisions are mine. AI is simply not up to the task yet. The real test is not when you can show what amazing things a self driving car can do - it's when you throw situations at it trying to make it do the wrong thing and you can't. No one wants to show those test yet, but those are the ones that matter. Chris -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.evdl.org/private.cgi/ev-evdl.org/attachments/20150406/d52546d5/attachment.htm> _______________________________________________ UNSUBSCRIBE: http://www.evdl.org/help/index.html#usub http://lists.evdl.org/listinfo.cgi/ev-evdl.org For EV drag racing discussion, please use NEDRA (http://groups.yahoo.com/group/NEDRA)
