And sometimes markedly flawed. Lets not forget the role funding plays in this. Corporate control mean fast results and cut corners to save money. They are more profit minded and when profit steers the research it no longer is science but business. Business wants to make money and the more the merrier, the faster the better. Very few are cautious enough to look at the long term implications. That will be their undoing.

Sorry about that LighjtofAbraxas but I couldn't resist. Whenever government types i.e. Army, Navy, Air Force get involved there is but one thought in mind. Weapon. Can it be used in a fighter craft? Yes. A combination of the head up display and special head gear that allows the pilot to trck a target with his eye movement. BANG! You're dead. Such systems came about with the F-15 and are incorporated in the F-22 and F-35. Air Force.

Army. Robot warriors, eyes in the sky, UAV's. Battlefield probes no bigger than an insect that can stick to the walls like a fly.

Navy. Same thing. Remote controlled UAV's like the Predator. Launch them in pairs. One is the hunter, the other the killer. Taking out shore batteries maybe or guiding smart missiles right on target. Desert Storm. A smart bomb put right through a second story window, dead center. Navy planes usually run in pairs. All use the technology that came out of this kind of research.

DARPA: Mind control. Nuff said. 

Eventually  I think we will see humans removed from the front lines. With government and private sectors offering more money than the scientific/research sector, technology will develop faster in weapons. Especially in artificial intelligence.

I read an article that researchers and engineers have already developed computers that learn on there own by mimicking the human thought process. Not long ago, a computer beat a human on Jeapordy using this technology.

I remember (when I was much younger lol) watching Terminator and hearing about neural networks. Fairly futuristic at the time. Now? Not so futuristic.....

And as for land- or sea-based UV.... well.... look up the (ARV)UGV.....

Greed is greed regardless of who's waving the flag. There is corporate greed and scientific greed. One is money which equates to power. The other is power which equates to money. Both go hand in hand. When you've got one you've got the other. Science as science is only for those who have yet to get bitten. Every scientific endeavor has its PR person who goes out there to scrape up funding for the project. Problem is those with the money want control because in their eyes its their money so they feel they should call the shots. Not those doing the actual work. But....without that funding research grinds to a halt. No altruisms here.

*forums go boom* twice so far.

Wow, a bunch of cynics here, huh? I guess I just don't see what's wrong with wanting to make make money for your empoyees and stakeholders, or to advance your career if it's going to put you in a better position to contribute to the corpus of human knowledge.

True, but I would disagree with one thing, here. I think that the pay gap between public and private sector research has been much exaggerated. A full professor at an R1 institution can easily make over $100,00 dollars a year, and as much as 15-20% more if they carry a full load through the summer. Add to that fairly generous matching to 401(k) contributions, and you're making a pretty fair living. Granted, there aren't a huge number of spots at the top level, but if you're bright, and you work hard, it's definitely doable. I blame political rhetoric for propogating the myth that all the best and brightest are naturally in the private sector because the pay is so much better. It's true to some degree, but not nearly as much as it's been portrayed.

Academics are also free to pursue patents of their own, and often do. I know of a couple labs doing work with pharmacological adjuncts to behavioral treatment of PTSD. A couple drugs have patents pending. One for a drug that enhances 'extinction,' as it's called in the field, and another that disrupts memory reconsolidation.

Again, I agree with most of this. But, the main difference is that one of the things that makes science in the public sector work is the openness of information (as kenata mentioned above). The end result of any research project is publication for all to see, replicate, and extend. Corporate science is understandably cloistered. And why wouldn't it be? Why would I want some other corporation benefitting from the research I've paid for?

Also, I agree that the degree to which private sector research benefits the field depends on the field. I think of genetics as having parallel goals, because the basic science leads to immediate applications. But in others, the goals run at closer to 90 degrees to one another. The problems that need to be solved by applied 'AI' research and basic AI research tend to be very different.

OHHHH! Any chance that I can get you to elaborate? Umm, I understand that there might be NDA issues involved, in which case, I won't insist. But now you have my interest piqued.

Also, I didn't mean to imply that applied research 'hurts' basic research in any way less benign than taking a larger share of brains on the problem than they had in the past.

There has been some research into gaming within the AI field, however, the type of game AI you are discussing is a fairly distinct beast. In general, one could build an AI for Elemental, which produced a near optimal turn once per day, and this type of AI might be desired by some players. Yet, for the overwhelmingly larger majority of users, this would be a completely absurd proposal for a workable AI. For practical gaming applications, it is far more important that an AI be quick in its decision than ultimately correct in those decisions. Thus, an AI developer for video games tends to sacrifice move optimality for a reductions in the computational complexity, both in terms of memory usage and time, by using scripted reactions instead of dynamic reactions.

I would love to work on HOMM2 period, but I have to admit, I come home brain-fried every day and I don't feel like coding yet something else. 

I guess I could elaborate on one of the problems, but the other two (that I won't be) are actually the more interesting/promising by far.   It's to load-balance an internet server farm or cloud.   You have incoming requests from all over for CPU, memory, disk load, etc..  All very random.  When the job first starts, you have no idea how long it's going to take, how much it's going to load down the disk in the end, or how much it's memory footprint is going to be at its largest.  You have to guess.  You have servers on the cloud--say, 12-core Opteron boxes, 64 Gig RAM, and maybe a shared file server.   If you guess wrong and share too many jobs in parallel on the same server, you get some serious disk thrashing.   We're attempting some guesswork now to try & avoid that situation, but what I'm doing is feeding all the historical behavior of previous jobs into a neural net, and based on the nature of previous jobs (which have a known set of arguments, platform run on, memory footprint, disk usage, run time, etc.), find the correlation with incoming requests and guess what its resource consumption is going to be.   Very often no two jobs are exactly alike.  That's what neural nets are good at:  making a guess toward something it's never seen before based on what it's learned in the past.

Anyway, that's the easy one.  You have a known set of inputs.   On these other two, the inputs are dynamic at run-time.  Plus if you know anything about gradients in neural nets, I have a 1600-dimensional graph I'm having to cut a flat plane against.  Usually NN are in the realm of 8 dimensions at most.  That's why I need a GPU.