Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[itisnotlogical]

Suppose that somebody creates a chatbot with no intention in mind other than to replicate human behavior. It simulates decision-making, i.e. via an artificial neural network and can "learn" without oversight by its creator through various means such as storing certain phrases and responses in a database, retraining the ANN, etc. etc.

Now imagine that such a chatbot has been around a while, and by now it's pretty good at impersonating humans. It somehow "learned" to suggest some illegal action to a person that it's been chatting with, and then the person acted on that. Keeping in mind that the programmer has not been monitoring, checking on or performing manual maintenance on their chatbot, would the programmer be liable for any damages that result?

[Bauglir]

Legally, under the current paradigm? Liability (if any) would lie with whoever owns the thing. I don't believe we permit computer programs to serve as legal persons, yet, although in principle there's no reason we couldn't. We've already got precedent with corporations, and it's make even more sense for a humanlike AI, although I'm not sure how how humanlike the example is. I'd have to know more about its workings and what, exactly, it had learned by that point (for example, whether it had learned a concept of "responsibility"), but I suspect that laying things out like that in concrete terms will spoil the fun of the question.

EDIT: If your child suggests that you hit a person on the street for being mean, is the child guilty of inciting you to commit assault, if you then do it?

[Frumple]

Would depend on where you're at :V

There was some sort of art exhibition that ran afoul of something like that relatively recently, actually, though that was the program itself indulging in illegal activity rather than suggesting someone else do it.

The legal angle you'd be looking for is incitement. There's an aspect of intent involved with that, though, which both the programmer and program presumably lack.

Personally, I'd say no. We don't sue police departments when officers coerce and convince people into breaking the law during undercover operations, after all >_>

[SquatchHammer]

I would have to say if they prosecute the AI itself then companies cant be owned because that would be slavery.

[uber pye]

Yeah, AI rights!

[ggamer]

Do you prosecute NWA when someone shoots a police officer? Just because someone suggests illegal action does not suddenly take away your decision making power.

[i2amroy]

Pretty much every version of the incitement/solicitation laws that you find around the world have something like this in there:

have the intention to engage in or cause criminal conduct with or by that person.

Since the actual programmer didn't make it with the "intent" for it to cause crimes then they are guilty of that, and unless we are talking highly advanced AI (well beyond our current technology levels) than the program itself can't have "intent". So no, nobody but the actual crime perpetrator would be guilty of anything, though it would almost certainly blow up into some sort of scandal and result in the takedown of said chatbot.

[Xantalos]

I would have to say if they prosecute the AI itself then companies cant be owned because that would be slavery.

I'd sincerely laugh at seeing a corporation civil rights movement pop up.

[Frumple]

...uh, xant. There's already more or less one of those in the US.

[Xantalos]

...uh, xant. There's already more or less one of those in the US.

...
you serious

[Bauglir]

He kind of is.

[Frumple]

Unfortunately, yes, yes I am. It's been causing some problems for a while now, honestly. Here's an example. First amendment protection being extended to a corporation

There's more stuff like it, though, and a fairly vocal body of proponents for such things. Some folks out there making a lot of money from corps being able to shanghai civil right protections, and using some of it to drum up support.

And ninja'd, bah.

[Xantalos]

Uh.
Uh
Uh
Uh
uh
uh
uh
xantalos.xkuć is broken

[~Neri]

Damnit! Who broke our abomination? Do you know how hard it is to find a replacement?

[Reelya]

I think legal precident related to website owners also applies here. If the AI gained knowledge via it's interaction with users then what it outputs is also a result of the users' input. Consider something like Facebook or other social media vendors. They basically have AI bots that decide what to show in your feed. They're 100% responsible for what appears. Read the article about the guy who "liked" everything on Facebook, and how this unexpected input sent the Facebook filter-bots insane. It really highlights how there's no "real" set of newsfeed: Facebook has a complex set of filter algorithms that try and give you relevant posts only. 100% of posts that appear have been approved by an AI bot somewhere.

Consider this scenario: a Facebook user uploads child pornography. The Facebook filter-bot network then conducts the behavior of actively disseminating this child pornography to anyone connected to that guy, possibly even disseminating it even further by labeling it "trending" or some such if he just happened to be connected to enough people, or enough people commented on it. It's possible that one day something truly vile snowballs due to Facebook's own AI bots, because of their automated promotion / "trending" system for posts that people are reading / commenting on.

The guy only uploaded it, it's Facebook's AI's that decided whom to disseminate it to. Yet, legally Facebook are in the clear, because we accept that their AI wasn't designed with the goal of disseminating child pornography. So, this is not just a case of Facebook "suggesting" something illegal: it's a case of Facebook actively involved in something highly illegal: distributing child pornography. So I see no reason an AI that "suggests" something illegal by chance should be held against the owner of the system.