The Steampunk Forum at Brass Goggles
July 22, 2018, 02:13:25 am *
Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
News: Support BrassGoggles! Donate once or $3/mo.
 See details here.
 
   Home   Blog Help Rules Login Register  
Pages: 1 2 [3]   Go Down
  Print  
Author Topic: Steampunk and Artifical Intelligence  (Read 3761 times)
hardlec
Snr. Officer
****
United States United States


Solutions do not need Problems


« Reply #50 on: November 17, 2013, 03:46:01 am »

There are problems that may neither be solved nor proven unsolvable.  There are also paradoxes that are more like parlor tricks than problems. 

The initial application for the OSS was designed to frustrate applicants into leaving is disgust.  If you sat through the whole thing, you should be an analyst.  If you stopped too quickly, you were to impetuous.

The test was designed to challenge emotions and character as well as intelligence.
Logged

Whatever happens we have got
The Maxim gun and they have not;
Technology is no substitute for Valor
Both are true.
MWBailey
Rogue Ætherlord
*
United States United States


"This is the sort of thing no-one ever believes"

rtafStElmo
« Reply #51 on: November 21, 2013, 06:02:04 am »

Actually, most mathematical theorems which are unprovable can be proven to be unprovable.

Of course there are the NP-complete problems which cannot be proven to be unsolvable, but since any NP-complete problem can be transformed into any other NP-complete problem, it is possible to prove that a problem is NP-complete.

Math is weird like that.



*Holds head in hands* I'm going mad, I tell you, M-a-d... Wink
Logged

Walk softly and carry a big banjo...

""quid statis aspicientes in infernum"
J. Wilhelm
╬ Admiral und Luftschiffengel ╬
Board Moderator
Immortal
**
United States United States


Sentisne fortunatum punkus? Veni. Diem meum comple


WWW
« Reply #52 on: November 21, 2013, 06:51:57 am »

Like others say before me, supercomputers are "just"  pseudo-AI. By brute force (speed) it calculates all possible outcomes and executes the right answer. It doesn't learn, it computes. I think there is a posibility in the organic world. Give some fast communicating collective (animal or bacterea) the opportunity to make strings of connections. Let it build the right combinations of strings, while ignoring or distroying the wrong combinations.

If I may ladies and gentlemen, I have observed that there has been no discussion at all on the differences between Artificial Intelligence and Computational Intelligence.

Mind you the sense of the terms Artificial Intelligence and Computational Intelligence which I refer to is not the same as it's being used in this conversation.  Instead, I refer to the IEEE definition of such terms.

Strictly speaking the IEEE defined Artificial Intelligence as being derived from calculations in mathematical logic, which absolutely requires the use of machine language on the form of symbols - we all know about binary numbers.  Arguments about the feasibility of solving a particular analytical problem belong to this definition, as Mathematics is really a human invention whereby we give substantive form to abstract thought according to some rules we have defined.  That is in fact the source of paradoxes and logical quandaries that cannot be solved.

However, the term Computational Intelligence covers an entirely different definition given to systems formed through the interconnection very large numbers of very simple analogue or quasi-analog/digital hybrid cells - more commonly known as Neural Networks, and whose architecture is what Mother Nature tends to follow.

Up until WWII the use of analogue circuits was commonplace for aeronautical controls, elevator controls, etc.  But most efforts were aimed at following simple known mathematics which could be derived from the laws of physics affecting electric or electronic components.  There are in fact analogue circuits which were originally developed specifically as calculators, first with vacuum tubes, and later with transistors called "Operational Amplifiers."  The name "Operational" refers to a mathematical operation , such as adding two voltages to make a third one which is the mathematical addition.  Again these are analogue and NOT digital calculations.

So engineers knew of different methods to process mathematical logic, other than using logical discrete gates, be it transistors or cogs.

Why do I mention this?  Because in some ways biological neural networks at the level of an individual neuron partly operate in such analogue fashion, and partly operate in a "digital fashion."  A single neuron may be connected to 1000s of other neurons and upon reaching a "threshold input" (the "digital part of a neuron's behaviour), it fires an analogue output proportional to the input levels.

What you can do with these networks is not only astounding, but actually forms the basis of all pattern recognition software, including speech, image processing and "fuzzy logic" type of solutions to mathematical problems.

What the public doesn't know is that while the neural networks could be built physically from analogue electrical components, it is more practical (albeit MUCH slower) to "simulate their existence in "virtual neural networks" which are actually coded in the form of (Oh! How ironic!) software for use in digital equipment (i.e it's easier to write a slow digital software programme, rather than physically build a fast minute network made of 1000s or millions of neurons - but make no mistake - an analogue silicon-based neural network would be much faster than the digital computer).

Moreover, this type of computational device is far more robust than a gate-digital device, in that it can sustain damage to a few of it's neurons, and as long as a minimum geometric pattern in the connections is maintained, the operation of the neural computer remains intact or nearly intact (that is why you can drink a few beers -  like I am right now- every now and then).

Last but not least, neural networks have to be trained.  In other words, neural computers are by default learning, and adapting machines.

And here I will stop.  Partly because I'm too busy burning some of my own neurons with this lovely Shiner Bock brown ale, and partly because I want to encourage people to perform their own research on the subject.

Keywords: IEEE, Artificial Intelligence vs. Computational Intelligence ~~ Neural Networks, Hopfield Networks, Feedforward Networks, Network Training, Fuzzy Logic ~~ Analog Controllers, Root Locus Method ~~ Cellular Automata

I've... seen things you people wouldn't believe..    All those... moments... will be lost in time, like [coughs] tears... in... rain.

Alright, alright I'm just being melodramatic. It's the beer talking Grin
« Last Edit: November 21, 2013, 08:06:24 am by J. Wilhelm » Logged

xanthra
Snr. Officer
****
United States United States


« Reply #53 on: November 23, 2013, 02:03:04 am »

mmmmmm Shiner Bock.  An intelligent choice : )
Logged
hardlec
Snr. Officer
****
United States United States


Solutions do not need Problems


« Reply #54 on: November 23, 2013, 06:18:38 am »

Don't believe anything I tell you.  Its a lie.

:-)
Logged
Pages: 1 2 [3]   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.20 | SMF © 2013, Simple Machines Valid XHTML 1.0! Valid CSS!
Page created in 0.317 seconds with 16 queries.