AI – Artificial Intelligence – PBP Challenge For The Fittest

[T.Jackson]

<h2 style="color: red" align="center">
AI – Artificial Intelligence – Past, Present & Future. </h2>

In early days, scientists thought that the computer would experience something like an “electronic childhood”, in which it would gobble up the World’s libraries and then begin generating new radical wisdom. Seldom people talk this today because the problem of simulating intelligence is far more complex than just stuffing facts into the computer. Facts are useless without the ability to interpret and learn from them.

Until recently, computer-based games of chess could normally always be beaten at the hand of an average human player. In 1997, the expert system dubbed "Deep Blue" was the first computer to win a chess game against World champion Garry Kasparov. Today, developing chess AI is no longer considered a challenge for scientists. Arguably, even some modern Windows-based games of chess can be set to almost normally always win.

Today, expert systems, which have the ability to carry out complex human tasks, are on the rise. An expert system is a software package that is used with an extensive set of organized data that represents the computer as an expert for a specific task. It mimics the decision making of a human expert in this field. These programs are daunting to develop.

But the question on most people’s minds still remains. Can a computer really think for itself? No, and it never will. A computer will never be able compose original music of its own or write like Shakespeare. A computer has no passion, emotion, feelings or original thoughts. Computers follow rules of logic and nothing ever stands between this.

There’s another big hurdle too. How in the World can a computer ever be programmed to understand natural languages? The human language is just too diverse. Often many like words and phrases have entirely different meaning. Developing a program that can distinguish between facts & questions using everyday human context seems impossible. Nonetheless, the study of this field continues and, at current, not looking very promising. <br/>

<img src="http://www.picbasic.co.uk/forum/attachment.php?attachmentid=1587&stc=1&d=117816665 6" align="left" >
<p>
Mid last year I developed an unbeatable AI for the well known game of Tic-Tac-Toe. A simple game indeed, but unfortunately not an equally as simple chore to code. After developing this project I gained some considerable respect for just how daunting AI programming can be. Note: screen shot shows an earlier version 0.3. A few trick moves allowed the player to beat AI every time. These have since been corrected.

Download game here (VB source code only): http://www.planet-source-code.com/vb...67151&lngWId=1
If you would like a compiled executable then this can be arranged.
</p>
<h2 style="color:green" align="center"><u>PBP Challenge For The Fittest</u></h2>
The challenge that I present to you all is to devise a game of Tic-Tac-Toe that is unbeatable using a graphical LCD, push buttons and other necessary components. Keep in mind that AI is only as strong as its creator's weakest link. Also, after a while, being beaten by a machine becomes quite frustrating, so the player may prefer a more intuitive opponent that isn't quite so perfect. In this instance AI will have "blonde" like moments where the player has an open space to make a move that could win them the game. This is a much more fun AI to play against. In my Windows version of the game I have coded AI to be about 95% unbeatable on hard, 75% on normal and 60% on easy. I confirmed that my AI is unbeatable by allowing the computer to play against itself many hundreds of times. Draw game always. And yes, Tic-Tac-Toe is indisputably flawed where best play leads to draw. <strong>Good Luck!</strong>
<br/>
Trent Jackson ©2007

[T.Jackson]

In addition - probably because I'm so open minded, I personally do believe it's possible for a computer to be programed to fully understand natural languages.
My personal opinion, I'm not saying this after reading from a text book - most books will probably share a similar negative attitude as I have already expressed.

[Darrel Taylor]

Trent,

A contest/challenge is typically associated with some sort of incentive.
Be it a prize, an award, or maybe just a gentleman's wager.

So assuming someone has spent the last couple days writing an Unbeatable Tic-Tac-Toe game in PicBasic Pro, and they decided to submit it.

What's the pay-off?
<br>

[T.Jackson]

I think a project competition with prizes would go down great. You already have a voting system in tact. But unfortunately, it's not such an obvious thing to be able to do. (This is one of the reasons why few threads actually have votes) The voting needs to be more clearly visible somewhere. It needs to be seen at a glance.

The other problem is that you're allowing people on the site without having to create an account. If you made it mandatory to sign up for membership after visiting the site more than a couple of times - this would possibly improve interaction. You could also enforce voting to some extent too.

[Darrel Taylor]

Sorry if that came out wrong!

What I meant to say was....

I've just spent the last two days writing an unbeatable Tic-Tac-Toe program for PicBasic Pro.

There seems to have been a challenge to do so, so I accepted it without even knowing Why. (other than it seemed interesting)

As you've already pointed out, it's not as easy as it first seems.
I'm currently trying to finish up the third version.
The first two did more toward teaching me how to play Tic-Tac-Toe than they did actually playing it.

One thing I found interesting with those two versions was, when you let the PIC play against itself, it was always a draw. But when you played it yourself, it was quite easy to beat. Even the first version was like that.

So I have to question the reasoning that letting a program play against itself somehow proves that it's unbeatable. It may just be playing at the same level.

With that in mind, I think that the only way to determine who's is the closest to Un-Beatable, we need to come up with a way to have the programs play each other.

But then there's that what's in it for me thing.
<br>

[T.Jackson]

Did you take a look at the Visual Basic source? If the protocol is correct - AI is solid if it always draws while playing against itself. But it can't always follow hard coded rules of logic. This I suspect is where your error is.

Protocol should be as follows:

AI checks for winning move.
AI checks to block opponents win.
AI makes move at random. (without this you can't verify AI's strength if set to play against itself)
<br/>