Slashdot Mirror
Computer Software to Predict the Unpredictable
Amigan writes "Professor Jerzy Rozenblit at the University of Arizona was awarded $2.2Million to develop software to predict the unpredictable — specifically relating to volatile political and military situations." From the article:
"The software will predict the actions of paramilitary groups, ethnic factions, terrorists and criminal groups, while aiding commanders in devising strategies for stabilizing areas before, during and after conflicts.
It also will have many civilian applications in finance, law enforcement, epidemiology and the aftermath of natural disasters, such as hurricane Katrina."
Go ahead and predict the weather for a week. I will be impressed.
Predict it for 2 weeks, I will blow you.
You cannot predict something with so many variables that you don't understand. You certainly cannot do it regarding how people will react.
Apparently there are those that have forgotten the old computer law of "Garbage In, Garbage Out". Even if we had a perfect model to predict these sort of things, we don't have any way of supplying the required data to model the prediction. What's the computer going to do, go undercover in secret groups? Read the web sites? Listen to radio chatter and analyze their conversations?
Maybe someday when we have a real science of A.I. something like this might be possible, but all it shows is that this university professor will happily take government money for delivering absolutely nothing.
Sometimes it's best to just let stupid people be stupid.
It isn't about actually being able to predict anything useful. Think of it like this. As a "World Leader" [sic], how much would you spend on the Ultimate Cop-Out(tm)? yeah a few million is a *bargain* for what this thing can do. None of the people involved in this project are actually interested in the predictions. What they are interested in is that the *next* time they have a royal screw up, they can say: "well, its unfortunate this happened, but you see, we have really smart supercomputer. It has 3-D and stuff. And it tells us what is most likely to happen. This wasn't on the list. We only have limited resources, and this is the best way to focus those resources where they are most likely to be doing us good".
Its the ultimate repudiation. As far as I can predict, they will spend lots and lots more money on this, get some buddies in on the gravy train somewhere to boot, and they still got themselves a bargain.
People who think they know everything are a great annoyance to those of us who do.
"...and the aftermath of natural disasters, such as Katrina."
Dealing with the aftermath of Katrina wasn't a matter of applying rocket science. It was simply a matter of simple logistics and a government that gives a shit about people. Unfortunately, the U.S. government has shown time and again under this administration that it could care less for the lives of its citizens, let alone the citizens of other countries. These problems can't be fixed by software. They can only be fixed by real leadership, something the people of the U.S. haven't shown much interest in electing...
It doesn't take software to predict that going into Iraq was a huge mistake. Just ask Chaney circa 1994. He knew it would be a major mistake, and he wasn't the only one. A lot of us were yelling and screaming to stop it before it started...
Software can't predict the future nor can they predict what stupid leaders will do. On Sept 10th, could anyone (or more importantly, any software) predict what things would be like in this country today? Even remotely? The war in Iraq, a country completely disconnected from 9/11. Guantanamo, spying on our citizens and other erosions of liberty... I doubt it. A single event and the responses by inept leadership led to a variety of disasters that nothing and nobody could have predicted.
You are, of course, presuming that Asimov was doing more than writing good FICTION. I sincerely doubt that any program will have any noticeable success in predicting anything, regardless of the wads of cash thrown at it to make it 'better'. Remember all the computers in the classroom priorities of the last few decades? How many of you used a computer in ANY bloody classroom that did not relate directly to the class? (ie- C+ programming {more likely FORTRAN}).
There's still the issue of dealing with the tail end of any distribution. I don't care about the 99.999% of people who, in the aggregate, fit a model. I care about that 0.001% of people who are going to completely blow it (because, as always, "past performance does not implly future performance").
A nice analogy when people think computers can make decisions or have "Artificial Intelligence" of any merit.
Pupil (Excited about AI):- I have just written a programme that writes music in the style of JS Bach.
Tutor (Seen it all before):- Really? How does that work then?
Pupil:- I programmed all of the known manuscripts by Bach and the computer uses that to write new compositions.
Tutor:- Great, can it write in the style of Mozart?
Pupil:- Sure, give me all the compositions by Mozart and I'll show you.
Tutor:- You mis-understand, can it decide, of it's own volition, to write in the style of Mozart.
Pupil:- Well, no it needs to base it's composition on something.
Tutor:- It has the entire works of Bach, is that not enough?
Pupil:- No, it needs the entire works of Mozart to write in the style of Mozart. Hell, even music students need to have heard Mozart in order to write in the style of Mozart.
Tutor:- Oh, so how did Mozart do it then?
America, Home of the Brave.