Forms of A.I. that should be invented

[PanzerTank]

I was wondering if you guys could conjure up some interesting forms of A.I. that have yet to be invented. By A.I. I mean anything from its simplest form as a pulley to a thermostat to a super-computer etc.

 

[kmguru]

We are working at it....Without data to work on, AI does not work very well. We presented an AI based fraud management program to one company who has a thousand Fraud Police but they were not interested.

 

[theoneiuse]

hey you know how much money you can make of this information
hope you get some some good info

 

[kmguru]

New ideas are hard to accept. But we are trying. We are working with the #3 defense contractor hoping they will have the right connections. They are very much interested. There is really nothing out there except a lot of hype. The fun is in building and solving a complex problem, money is a side benefit....right!

 

[weed_eater_guy]

oh wow, i didn't know you were into this,

what about the idea of a neural net? where small programs work together to simulate neuron interaction rather than a large program try to simulate intelligence? it sounds easy enough, why hasn't it turned into a powerful AI? just wondering if you could shed light on that, it'd be awesome if you could

 

[kmguru]

Yes, small neural nets would work, because AI has a predictive component. The reason you do not see AIs or real advanced AI type applications out there is because....it would require a lot of money and some serious architecture and programming to create the next generation program. FBI will be spending $475 million dollars to do their case file system called Sentinel. They could easily do such a program...but the prime contractor Lockheed Martin does not have the expertise to add such a ground breaking AI system. No one wants to take the initiative because they are afraid, it could fail...

and imagine, if after spending say $50 million, it does not do what you want to do...it would be bad news. Congressional enquiries and so on...So....you will not see such applications anywhere for a long time....not in DOD, nor in DHS or anywhere....

In AI, there is no such thing as incremental application...either you put the lego blocks together to make one or you do not...no half way that will provide any real value...and it will cost money to put it together.

 

[weed_eater_guy]

i had the idea that you could just make more programs and tell them to sort out their own architecture, where sheer quantity of "neuron" programs equals increased intelligence. i didn't think about the idea that they might need an initial architecture, is this the case?

and "next generation program", that means that there are set-in-stone levels of AI? or is it just next-gen as an industry term for a new set of goals? pardon the questions, i just find this field fascinating.