The History of AI

[bert]

Explain what contribution to Artificial Intelligence is the most important and defend your argument.

[woogley]

wow, is that a conversation starter or an essay assignment?

[darkprophet]

Well, the most notable was the creation of a computer, and to be honest, i dont feel like defending today, id rather flank the opposition using some pathfinding and attack with the horsies and ponies and goats. Then, neural network of the computer will no doubtly send in the seaguls, so you better have some anti-pecking armour on those goats.

Oh wait, that wasn't the question...

DP

Games published by our own members! Check 'em out!

[kevglass]

Goats? Seriously? - battle goats? with rocket launchers (guided missles of course - hence being on topic in AI).

BATTLE GOATS TO THE FORE!

Kev

[beowulf03809]

Forget anti-pecking...as anyone will tell you that has lived near a gull-infested location you had better have a big umbrella or you will find yourself in charge of a large force of smelly, white-washed goats and horses. 

[CommanderKeith]

What is AI in games?  It seems to be used pretty loosely because most of the AI we do in games is not Intelligent at all. 

Computer 'players' just do what they do in an optimal manner and don't actually learn anything.  To make up for their lack of over-arching strategy they are made flawless tacticians.

evolutionary AI and pattern recognition sound reasonably intelligent, but most other other things seem to be just maths optimisation.

[bert]

It's not an essay question... I just want to know other people's opinions! My favorite contribution was the tracking AI used in ROBOTRON: 2081.

if(robot.x>player.x) robot.x--;
if(robot.x<player.x) robot.x++;
if(robot.y>player.y) robot.y--;
if(robot.y<player.y) robot.y++;

It's really simple, summed up in four lines of code but IT WORKS and I think the guy who came up with it was a genius.

On another note:

CommanderKeith, you might be interested in this article about evolutionary AI:

http://www.popsci.com/popsci/science/0e13af26862ba010vgnvcm1000004eecbccdrcrd.html

[CommanderKeith]

Great article, thanks for posting it  .  I thought it wouldn't hold my interest for long enough to keep me reading but yes I read all 3 pages.

It made me think what evolutionary AI depends on is what kind of parameters or algorithm that it uses to evaluate itself (or its prodigy).  Surely there is a place for economic concepts of utility to come into that... I don't imagine economics would have been combined considering AI is typically a computer science problem.

[bert]

It is pretty creepy if you think about it, because then the system could be adapted to manufacture parts, building new drives and such, then because it's evolutionary it would continue to improve itself ON ITS OWN to take over the world. It gets a little scary, but I think that we'll never achieve anything of that scale. It all seems way too sci-fi.

[oNyx]

>What is AI in games?  It seems to be used pretty loosely because most of the AI we do in games is not
>Intelligent at all.

Heh. Well, at least its artificial

Most AI (even good one) in games is usually just good scripting. The thing is it only needs to look human like (result wise)... so, you can usually take a lot of weird non-human alike shortcuts. Like perfect aiming + some random error instead of natural less-than-perfect aiming, which would be based on lots of weird factors like perspective, (degenerated) experience, caffeine level, non perfect inputs, distraction factors... but at the end of the day its pretty much identical to the cheap solution.

For a pong ai you can for example predict the position where the ball will be... apply some error (which gets smaller the closer the ball gets), acceleration/deacceleration, reaction time... this kind of things. Its pretty human like, but its just a short script. But thats fine. Really. It only needs to look sorta plausible.

Games published by our own members! Check 'em out!

[arne]

In short: The AI has to pass the "Turing Test" - how it does that, doesn't matter. And with that we get back to history - that test is OLD - actually it's from the time, when Computers where in their children shoes.

[jfelrod1960]

Explain what contribution to Artificial Intelligence is the most important and defend your argument.

Soft computing methodologies like Fuzzy Logic, Neural Networks and Evolutionary Algorithms.  They have changed our way of life.  Fuzzy Logic is used in cruise control systems in automobiles, automatic washers in homes for example.  Neural networks are used everywhere.  Too many to mention.  Evolutionary Algorithms are still in its infancy but making progress.

As far as games is concern I don't know.  I have been working "off and on" on an AI engine for quite a while and I'm focusing on "real time learning" where the game object learns during play.  Not an easy task at all.  But I try.

[CommanderKeith]

As a digression, what do people think about an evolutionary algorithm to snoop out share market inefficiencies (read: money making strategies)?

I'm studying undergraduate finance down here at uni and I've been offered a place to do a PhD in Finance.  All I need is a topic, and I think that evolutionary AI applied to the equity markets would be interesting enough to fill up 3 years.  There's some work that's been done on it already, but I've got a feeling the finance academics wouldn't have the programming literacy to really look into the issues properly.  Maybe I wouldn't either!

[beowulf03809]

That would be a very interresting study, Keith.  Not only for study's sake, but also have some significant potential.  For example:

1.  You can develop a market-modeling game for the die-hard finance geeks.
2.  When it's done you can pattent it and sell a kit on infomercials for 19.95 that will let people use it to help their home day-trading business.
3.  You can keep it all to yourself, use it to predict market fluctuations and self-invest your way to billionaire status.

Either way, those all look pretty win-win to me.   

[CommanderKeith]

Thanks beowulf, another important reason would be

4. Delay having to get a real job for another couple of years. 

[MasterNag]

As a digression, what do people think about an evolutionary algorithm to snoop out share market inefficiencies (read: money making strategies)?

I'm studying undergraduate finance down here at uni and I've been offered a place to do a PhD in Finance.  All I need is a topic, and I think that evolutionary AI applied to the equity markets would be interesting enough to fill up 3 years.  There's some work that's been done on it already, but I've got a feeling the finance academics wouldn't have the programming literacy to really look into the issues properly.  Maybe I wouldn't either!

I know that there's already been a working project, few years ago, which was used to model the French Stock Exchange (Paris-Euronext) using genetic algorithm and various AI technologies... If interested, feel free to contact me, as I have little knowledge in this field for having investigate it a little

[CommanderKeith]

Sounds interesting.  I sent you a Personal Message.  Looking forward to hear about it!

[blahblahblahh]

It is pretty creepy if you think about it, because then the system could be adapted to manufacture parts, building new drives and such, then because it's evolutionary it would continue to improve itself ON ITS OWN to take over the world. It gets a little scary, but I think that we'll never achieve anything of that scale. It all seems way too sci-fi.

Nah, the problem is that the article carefully hides the truth of EA - its hidden in this simple phrase:

"It then analyzes them in KOJAC, which assigns each a fitness rating based on how close it comes to a desired set of specifications"

Inventing a fitness function is *VERY HARD* to do, and is why EA hasn't taken over the world. A bit like hitchhikers guide to the galaxy - the greatest computer ever built solved the answer to life, but had to build an even bigger computer to work out what the question was - a computer would have to be incredibly intelligent in order to work out what the fitness function were that would produce something cleverer than itself. At least, thats the way things stand with current knowledge.

[bert]

Glad I started such a hot topic!

[Raghar]

In short: The AI has to pass the "Turing Test" - how it does that, doesn't matter. And with that we get back to history - that test is OLD - actually it's from the time, when Computers where in their children shoes.

Turing test isn't mathematical test, because it depends on experimentator, mainly on his cultural background. Some people didn't passed it.
 

[Raghar]

It's not an essay question... I just want to know other people's opinions! My favorite contribution was the tracking AI used in ROBOTRON: 2081.

if(robot.x>player.x) robot.x--;
if(robot.x<player.x) robot.x++;
if(robot.y>player.y) robot.y--;
if(robot.y<player.y) robot.y++;

It's really simple, summed up in four lines of code but IT WORKS and I think the guy who came up with it was a genius.

With excepion that one if is missing. And if they shouldn't occupate the same place it should be > (a + 1)

[ENC]

Hmmm do you think the gamework RoboCode is good for AI developing? Just asking.. cause I have seen many RoboCode Compeitions and they came up with many good bots

[kappa]

RoboCode is ok, tried it a long time ago, don't think its that good but for a little ai understanding it ok, didn't like the way they over abstract the java language.

btw this is a pretty old topic you just resurrected and you question isn't really relevant to it, i've noticed you do this quiet a lot, it would be better-nicer if you started a new topic instead unless your question is directly relavant to an old topic.

[ENC]

Ok Cheers!