Artificial......Intelligence?

I like NCIS. But sometimes, I wish it were better.

The episode I saw the other night, which we'd taped last week (possibly two), and which was a two or three year old rerun, involved a homicide in a lab building housing an autonomous AI-controlled Hummer. A test engineer, sitting in the vehicle, dies when the vehicle locks its doors, hyper-snugs up her seatbelt, and then pumps carbon monoxide fumes into the passenger area. The NCIS people (except for their resident geeks) get to use the phrase 'robot car' a lot , and express befuddlement and amazement many, many times. Wow, this is the same CPU as they use on the Mars Rover! Determining that the AI system is massively complex, so that they would not know where the code was that did the nefarious deed if it was pointed out to them, or even if it was the AI at all (this was what I was hoping for), they look elsewhere, and discover that a piece of software was inserted into the boot sequence of the AI; said new software set up the AI to do the nasty. They conclude it must have been added via a 'chip', and miraculously find it almost immediately, hidden in the gas cap.

Then the show starts to get implausible. They rapidly figure out who did it, and come up with a clever and possibly even legal way to trap the miscreant. Wrap, and go to commercial.

I don't know buckets about AIs, but I know a little, just as I know a little about autonomous vehicles -- mostly from reading about the annual DARPA challenge. So far as I could see, nothing that they mentioned was inherently implausible. Well, the carbon monox, perhaps -- even autonomous vehicles don't usually include a provision for venting exhaust into the cabin. And the idea that two good-guy geeks could successfully find all this, hey, presto -- well, possibly. With a lot of luck. But that the guy who set up the software did it because he wanted to kill the lead engineer because said engineer owned a lot of stock in the company underwriting the build, and if they were successful in their drive-off test, would have owned the technology, which isn't tolerable, so, hey, its not personal, just business? Then a miracle occurred.

I know that it's asking a lot for a TV program to realistically show a malevolent AI -- even one that is not sentient (I say that as if why of course they are!), but one where, simply, the system doing reasonable, expected things had an unreasonable, unexpected end. That would have been fun. That would have been interesting. The plot solution they chose was, in comparison, quite pedestrian.

Maybe next time.