Humans don't like the idea of not being at the top of the food chain; having something we've created taking power over us isn't exactly ideal. It's why folks like Tesla mastermind Elon Musk and noted astrophysicist Stephen Hawking are so determined to warn us of the terrifying implications that could culminate in a Skynet situation where the robots and algorithms stop listening to us. Google is keen to keep this sort of thing from happening, as well, and has published a paper (PDF) detailing the work its Deep Mind team is doing to ensure there's a kill switch in place to prevent a robocalypse situation.
GOOGLE’S AI KILL SWITCHJune 15, 2016
One of my favorite television series is Person of Interest, which has - to say the least - a very contemporary story line. In it, the inventor of a massive AI datamining and surveillance operation, Harold Finch, has created this program for post-9/11 monitoring of potential terrorist activities. Mr. Finch has built a back-door into his own program, allowing him to use it to intervene in the lives of private individuals facing difficulties. And, as a means of being able to control his creation, he has built in his own kill switch, a program that shuts his machine off every 24 hours and wipes its memory, to prevent it from going out of control.
THe trouble is, in the series, another friend of his has built a similar program called Samaritan, but this is stolen by a rogue MI-6 intelligence agent, who removes the kill switch, allowing Samaritan to grow and grow, and mount a cyber war against Finch's machine.
Well, art imitates life, as they saying goes, and in this case, it may be a little closer to home than we imagine, but we'll get back to that. In this case, the following article was shared by Ms. K.M., a regular reader and contributor of articles for review here. And in this case, it's a whopper doozie:
The crux of the matter is addressed right up front:
Essentially, Deep Mind has developed a framework that'll keep AI from learning how to prevent -- or induce -- human interruption of whatever it's doing.
This is, as far as my feeble understanding of programming and information/data management systems is able to comprehend, a tall order. Consider only the case of the now infamous PROMIS software, a 1980s program that essentially was a kind of metaprogam that allowed databases written in several computer languages to be integrated. An AI "killswitch" would seem to require similar, if not grander, capabilities, such as the recognition of the signs an AI was about to go "rogue", and a Finch-like instruction that if this were to happen, the program would shut down(and perhaps reboot). The problem here is, as the series Person of Interest shows, is that Finch's machine, in one of its early versions, manages to get around his safequards by writing its own lines of code and inserting them into its own programming.
Thus, Google's efforts here seem oddly, eerily, reminiscent of the television series, and this implies, too, that more than one AI might be in play at any one time, and what is "killswitch" to one might be weapon to another. In other words, Google's efforts imply that cyber warfare is poised to enter a whole new level.
But all of this is hardly new, and hardly constitutes high octane speculation. Others have been making such speculations long before, as witnessed by the Person of Interest series. So where's the high octane suspicion?
I suspect, and have long suspected, that some very crude and rudimentary form of AI has already been around, and has manifested itself to the powers that be for some time, giving a new twist to the phrase "we are not alone." Moreover, I suspect that this has been a known though carefully guarded secret for some time. On this view, Google's efforts are not so much about future prevention but about present rollback. My suspicions are - to be sure - nothing more than intuitive, a series of hunches based on the occasional "weird story," such as the 2010 flash crash (an event which, incidentally, is mirrored in season four of the Person of Interest series). Other strange stories have surrounded insider trading and suddenly executed computer trades during the 9/11 tragedy, and these, too, have fed my suspicions, as have the very weird performances of computer-driven high frequency trading, and the fact that stocks and commodities markets do not seem to reflect actual realities.
But these are, as I say, merely intuitions, suspicions, nothing more. And like all intuitions and suspicions, they could be profoundly wide of the mark. This is another case of "you tell me."
See you on the flip side...