In other assorted craziness:

http://www.boingboing.net/2009/11/02/mechanical-computer.html

It doesn't explain it very well, but this is the very first AI algorithm 
I ever saw. Mental, eh? A bunch of matchboxes can learn to play 
tic-tac-toe. Unless it loses its marbles...

(Of course, the intelligent commentator will realise that it's actually 
the human operator that is learning, not the matchboxes.)

Post a reply to this message

On 12/05/2010 12:29 AM, Invisible wrote:
> (Of course, the intelligent commentator will realise that it's actually
> the human operator that is learning, not the matchboxes.)

No.

The point is that the human operator does not apply intelligence other 
than following a mechanical process.  A sufficiently sophisticated wind 
up device could execute the algorithm.

The change in the box states is what 'learns'.  This happens by removing 
possibilities that lead to a loss.

So it *is* the matchboxes that learn.

Post a reply to this message

Paul Fuller wrote:

> So it *is* the matchboxes that learn.

More exactly, it's the algorithm for changing the contents of the 
matchboxes that learns. Not the matchboxes themselves.

Post a reply to this message

Invisible wrote:
> Paul Fuller wrote:
> 
>> So it *is* the matchboxes that learn.
> 
> More exactly, it's the algorithm for changing the contents of the 
> matchboxes that learns. Not the matchboxes themselves.

No, because the algorithm is the same at the start of each game. The 
algorithm *teaches*.  The matchboxes *learn."


On a side note, I saw a fun idea for an AI game a long time ago. In a 
similar vein, the program knew the legal moves and what constituted a win or 
loss, and would enforce that. The program would play based on this idea of 
wins add probability and loses subtract probability. But the human player 
wasn't told the rules. The game was whether you could figure out the rules 
and strategies before the game learned well enough to be unbeatable.

-- 
Darren New, San Diego CA, USA (PST)
    Ada - the programming language trying to avoid
    you literally shooting yourself in the foot.

Post a reply to this message

if there's one thing the movie War Games taught me is never to teach 
games to a computer, specially Tic Tac Toe!!

Invisible escreveu:
> (Of course, the intelligent commentator will realise that it's actually 
> the human operator that is learning, not the matchboxes.)

if you replace the human operator by a scanner to make matches and 
robotic arms to get beans out/in to the matches, it's the "system" that 
learns.  Though it probably has no clue that it is "learning".

-- 
a game sig: http://tinyurl.com/d3rxz9

Post a reply to this message

On 5/11/2010 9:29 AM, Invisible wrote:
> In other assorted craziness:
>
> http://www.boingboing.net/2009/11/02/mechanical-computer.html
>
> It doesn't explain it very well, but this is the very first AI algorithm
> I ever saw. Mental, eh? A bunch of matchboxes can learn to play
> tic-tac-toe. Unless it loses its marbles...

Now there's someone with too much time on their hands .... ;)


-- 
~Mike

Post a reply to this message

>> It doesn't explain it very well, but this is the very first AI algorithm
>> I ever saw. Mental, eh? A bunch of matchboxes can learn to play
>> tic-tac-toe. Unless it loses its marbles...
> 
> Now there's someone with too much time on their hands .... ;)

In fairness, at the time a real computer would probably have been 
*slower*...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On 12/05/2010 3:17 AM, Darren New wrote:
> Invisible wrote:
>> Paul Fuller wrote:
>>
>>> So it *is* the matchboxes that learn.
>>
>> More exactly, it's the algorithm for changing the contents of the
>> matchboxes that learns. Not the matchboxes themselves.
>
> No, because the algorithm is the same at the start of each game. The
> algorithm *teaches*. The matchboxes *learn."
>

I'm getting the impression that 'Invisible' is a large pile of match 
boxes filled with random numbers of beans.  Every now and then he/it 
makes an interesting observation followed by often incorrect statements.

We are gradually teaching he/it about the Universe.

One day true match box intelligence will rise up.  I expect it to be 
speaking Haskell.

Post a reply to this message

Paul Fuller wrote:

> I'm getting the impression that 'Invisible' is a large pile of match 
> boxes filled with random numbers of beans.  Every now and then he/it 
> makes an interesting observation followed by often incorrect statements.
> 
> We are gradually teaching he/it about the Universe.
> 
> One day true match box intelligence will rise up.  I expect it to be 
> speaking Haskell.

By that definition, it might have risen up already. Speaking Haskell, 
who would know the difference? :-P

Post a reply to this message

nemesis wrote:
> Though it probably has no clue that it is "learning".

'course not. If it had clue of that, it'd be called "self-awareness", and 
you should probably be running for your lives.

Post a reply to this message