THE PENTAGON’S PREDICTIVE AI: A GLANCE BACK
Now, before you remind me that I've already blogged about this story a few years ago, let me explain my reasons for revisiting it, via this version of it that was shared by T.M. (with our thanks). Briefly put, the recent blog I did about artificial intelligence becoming locked into chains of "thinking" that eventually lead it into complete nonsense or "model collapse" is the context for this "revisit." (Q.v. the blog
For many years I've been toying with a scenario that somewhere there exists a super-computer with a big super-computer program called the "World Options Gamer" (W.O.G.) or "World Utility Sentient Scenario" (W.U.S.S.) or some such grandiose name. The program analyzes massive amounts of statistics, and the computer generates so much heat it requires a cooling reservoir the size of Lake Erie to keep the computer from meltdown. On a regular basis, however (say every week), the computer composes and spits out its "World Homeland Analysis Matrix" (W.H.A.M.) for Amairikuh's Chief Executive Oaficer and cabinet, armoire, and the other furniture taking up space in the White Out House. On the basis of these weekly W.H.A.M.s, the chief executive oaficer and his armoire have been executing policy around the world and within the country that they themselves have not formulated. And because they have been irretrievably enstupidated (along with the rest of the country), they are no longer capable of formulating anything, including shaking a bottle of baby formula.
Something very much like this scenario I strongly suspect lurks between the lines of this now three-year-old article. Bear the W.O.G. with its weekly W.H.A.M. in mind as you read these sentences:
The US military's AI experiments are growing particularly ambitious. The Drive reports that US Northern Command recently completed a string of tests for Global Information Dominance Experiments (GIDE), a combination of AI, cloud computing and sensors that could give the Pentagon the ability to predict events "days in advance," according to Command leader General Glen VanHerck. It's not as mystical as it sounds, but it could lead to a major change in military and government operations.
The machine learning-based system observes changes in raw, real-time data that hint at possible trouble. If satellite imagery shows signs that a rival nation's submarine is preparing to leave port, for instance, the AI could flag that mobilization knowing the vessel will likely leave soon. Military analysts can take hours or even days to comb through this information — GIDE technology could send an alert within "seconds," VanHerck said.
We're "assured" that "The platform could be put into real-world use relatively soon." And its benefits are phenomenal:
The AI is looking for out-of-the-ordinary clues, such as a greater number of parked cars or aircraft. It can't say for certain what's happening — humans will still be heavily involved. Even so, the tech could be worthwhile if it prevents a 'surprise' attack or leads to negotiations instead of conflict.
Now we all know that this is where the precise danger lies. Assuming that "they" will be successful in creating this monster, who is to say that the monster will not try to lock the human analysts and decision makers out of the loop altogether, and maybe even "arrange" things to be able take matters into its own hands? We've been warned about such a scenario for years, from authors as diverse as Robert Heinlein to Isaac Asimov, and who can forget the Matthew Broderick movie from the 1980s, Wargames, where a super-computer called W.O.P.R. inside Norad's Cheyenne Mountain Complex decides to launch all of America's nuclear missiles at the Soviet Union because it's the logical thing to do (take them out before they do the same to us). I can see it now: the W.O.G. or the W.U.S.S. detects massive concentrations of Muslims in Arabia swarming around Medina and Mecca for the annual Haj, and launches swarms of drones as the Chief Executive Oaficer and his armoire are trying to find the off switch or the electrical plug.
But the real dangers are much less hypothetical and much closer to realization than one might think.
The most disturbing thing about the article is that the deployment of such an AI is simply assumed, it is necessary to be able to sort through all the massive amounts of data that "decision makers" need to see, and it's necessary because...well.... everyone else will do it. So here's the problem: that sorting algorithm is only as good as the programmers programming it. This means, like it or not, that the W.O.G. and the W.U.S.S. are theoretically open to being hacked, to nasty lines of code inserted during the construction of the program, or at a later point. But more importantly, those "sorting" algorithms will be subject to a cultural selection bias: a Chinese or an Iranian programmer will have subtle cultural influences - ones he is perhaps not even aware of - guiding the way selections are made. The problem is similar to that of the computers that review the billions of particle interactions taking place in CERN's Large Hadron Collider, because it is the computers that make the first sorting pass through billions of particle reactions to bring to the attention to the human analysts and scientists. But the sorting algorithm is based on a physics bias of what physics model the programmers of the computers had in mind when they programmed them; which model did they use? The Standard Model? Loop Quantum Gravity? Were the computers programmed to look for indicators of torsion effects? or not? And so on. And ultimately, who is to say that there is not a "secret" filter (or several such filters or sorting algorithms) designed to pull any anomalous data that those scientists and programmers wish to keep secret, and to shut that data to secret committee for analysis?
Nothing.
Thus, such "World Options Programs" are sorting programs, and consequently very powerful tools to control data flows and keep certain things from the public...and even from the secrets keepers themselves.
... and even from the analysist themselves.
Flash crashes? "fact checkers? internet censorship? You aint seen nothin' yet...
See you on the flip side...
(If you enjoyed today's blog, please share it with your friends.)
No comments:
Post a Comment