Ethical Artificial Intelligence

[Donks]

Well, producing ethical laws of robotics is hardly a new idea. Asimov got it down to three, though.

And it's not like robots generally follow those 3 laws, much less his proposed 10 laws.

ETA: Just found that he's been pimping his patent since at least October of '03, which generated pretty much the same reply as mine: Do you have results? To which of course the answer was *crickets*.

E2: Another spaming, this one from June '04. The exchange is interesting because, just as here, his replies don't actually follow from the posts others made. He first posts the OP, and whatever the reply, he posts the thing with the 10 laws.

E3: And yet another spaming, this one from Feb '05. Longer thread. And someone posts a great link to a blog entry which pretty much sums up my feelings on his patent. The best part is the begining:

Some goon (sorry: Californian counsellor) has patented Inductive Inference Affective Language Analyzer Simulating Artificial Intelligence (including the Ten Ethical Laws Of Robotics). It’s nothing but unintelligible babble, interspersed by (inaccurate) references to artificial intelligence theory. The author (who also writes a book on family values with a distinct evangelic slant, from which most of the text of the patent seems to be taken) appears to know nothing about A.I. or computer science.

And with that, I hereby leave this and all other spaming of this so called "breakthrough."

E4: Ok, I said I'd leave, but this is just too fun. There have been loads of threads slamming this patent since it appered. For instance this one.

Just when you think the US Patent office has reached the pinnacle of stupidity and irresponsibility, they suprise you by achieving a new extreme of insanity. Move over software patents, gene patents, and business method patents. Now you can patent systems of ethics. In July, 2003, a patent was granted for "The Ten Ethical Laws of Robotics" to John E. LaMuth, a family counselor and author of self-help books on ethics. LaMuth's "holistic theory" of ethics reads like a mix of greek philosophy, freudian psychology, and new-age psychobabble.

Asimov's first law requires at least some definition of what a human is, what an action is, and what sort of actions might harm a human. But it's probably a within our reach to build a machine that could make some good guesses. On the other hand, I'm not even sure what LaMuth's law means, so how is a robot going to figure it out and obey it? What is "nostalgia" with respect to a robot? What is ego and how would the robot determine if it had one, much less determine how to "express its individualism" within one of its "states"? And that's one of his more comprehendable laws. Later ones degenerate into new-age religious talk requiring the robot to support "ecclesiastical traditions", the "spirit of ecumenism", and to "profess a sense of eclecticism".

Asimov's laws act more like a safety on a gun or the guard a power tool - they were just intended to prevent humans from harming themselves with an intellgent tool they'd created. (and they're fictional, Asimov created them to move his plot along in a story, not to use in real robots).

LaMuth's laws look like the result of sloppy thinking by a non-technical person. They are what Douglas Adams would call a load of dingo's kidneys.

 

[Beerina]

(and they're fictional, Asimov created them to move his plot along in a story, not to use in real robots).

He even at one point introduced the concept that robots had been built with this for so long, that it was so ingrained in their design, that it could not be removed without extreme difficulty.

Which itself seems silly, but had to be done or there'd be millions of warbots constructed.

 

[jay gw]

A patent application for a temporary patent is only about 50 dollars and no you don't have to prove anything to get it.

Wouldn't the first place for a tech investment be a venture capital firm?

 

[cyborg]

One only has to look at some of the trivial patents made in the software world to realise the simple fact of the matter: patent clerks just don't have a clue.

 

[Beleth]

The main question patent clerks are there to answer is "Has anyone else patented this idea before?" Remember, a few years ago someone got a patent for a particular method of swinging in a swing...

A patentable idea doesn't have to be useful, or profitable, or even feasible*. It just has to be original. And I'm sure Mr. LaMuth's system is original.




[*] I think perpetual-motion machines need to be feasible (i.e. have a working prototype) before they are patented. Which is why there aren't any patents for them.