1. RE: The A.I. Project II
> > Start with worm instinct, get a full-potential worm.
> > Start with monkey instinct, get a full-potential monkey.
> > Start with human instinct, get a full-potential human.
>
> I haveto draw the line there. Some of us are "full potential
> humans", and are discarded.
Kat, are you saying you've reached your full potential?! I find that hard to
believe. As long as you have a brain and it can learn, you've not reached
your full potential, right?
Every human being has flaws. What if we could "create" a human without
flaws? Probably impossible...
> Yet humans want to build electronic clones of humans, to do
> what we could be doing, but aren't allowed to do.
The more one wraps one's brain around AI, the more one realizes that without
an exponential leap in hardware processing capability, we ain't gonna
realize the dream for another hundred years or so... or 40 years, as that
one website hypothesized.
2. RE: The A.I. Project II
> As far as finding out what is instinctual for a worm,first of all
> that would entail finding out what kind of worm we are talking about,
> and then doing research on that particular kind.
I think it'd be easier to invent our own virtual species of worm that lives
on a PC.
The problem is, we want to create a sentient entity, right?
> And thats not to say
> we couldn't just invent a type of worm that has instincts that we
> provide.After all, we will become,in effect,its creator!
Probably the easiest option, wouldn't you say?
There are two approaches:
1. create an actual AI entity whose "world" is the PC (+internet). The PC's
hardware thus becomes its "body."
2. EMULATE an AI entity in an artificial (emulated) virtual environment (a
worm in a glass bowl, for instance).
I don't know which would be easier. I'd like to hear thoughts on it before I
delve into it more.
3. RE: The A.I. Project II
> When you talk about visibility,in my mind your talking about graphics,
> Im sure we can come up with some 2d stuff that looks great.
When I talk about the AI entity's "visual" and "aural" abilities, I'm
talking hardware components. You would have to answer, "How does this
virtual creature 'see?'" "How will this virtual creature 'hear'?" Obviously,
you'll have to have graphic and sound processing code (code to analyze a
"picture" and be able to make distinctions in what it finds; code to take
noise and find relevance somewhere), which already takes this project beyond
a short-term exercise in fun. :)
4. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 06, 2002
Maybe we should start with the very basics first.
Maybe get some 'brain base' to be able to play tic-tac-toe and sort of
learn, then move on to slight more complex games like neutron and some
other simple ones, then work on expanding it.
5. RE: The A.I. Project II
> Maybe we should start with the very basics first.
>
> Maybe get some 'brain base' to be able to play tic-tac-toe...
Then you're simply making an expert system that can play tic-tac-toe.
And playing tic-tac-toe is no easy feat. You say that's the basics, but
you've gone waaaaay beyond the basics to suggest that our AI entity can play
a game and learn from it!
I'm thinking just a virtual worm (of our own design) in a virtual bowl of
dirt where we sprinkle some water and food every once-in-a-while and see
what happens.
To be truly AI, you have to think "primitive." Nay, "very primitive." :)
6. RE: The A.I. Project II
I'd like to see some information, tutorials, or libraries that can use
simple neural nets. That's the sort of thing I want to get involved in, but
I don't know much about them, or how they are implemented.
=====================================================
.______<-------------------\__
/ _____<--------------------__|===
||_ <-------------------/
\__| Mr Trick
>From: dm31 at uow.edu.au
>Reply-To: EUforum at topica.com
>To: EUforum <EUforum at topica.com>
>Subject: RE: The A.I. Project II
>Date: Thu, 07 Nov 2002 14:18:44 +1100
>
>
>Maybe we should start with the very basics first.
>
>Maybe get some 'brain base' to be able to play tic-tac-toe and sort of
>learn, then move on to slight more complex games like neutron and some
>other simple ones, then work on expanding it.
>
>==^^===============================================================
>This email was sent to: mistertrik at hotmail.com
>
>
7. RE: The A.I. Project II
On 6 Nov 2002, at 21:00, C. K. Lester wrote:
>
> > > Start with worm instinct, get a full-potential worm.
> > > Start with monkey instinct, get a full-potential monkey.
> > > Start with human instinct, get a full-potential human.
> >
> > I haveto draw the line there. Some of us are "full potential
> > humans", and are discarded.
>
> Kat, are you saying you've reached your full potential?! I find that hard to
> believe. As long as you have a brain and it can learn, you've not reached your
> full potential, right?
Wrong. I can feel my abilities slipping, and it's a proven fact that living in
total
isolation from society will cause what i feel: slower wit, less memory, and
apathy about everything. I go 3 weeks routinely without laying eyes on
another human, much less speaking to a human in real life, unless i goto a
store to buy something. So yes, i feel like i am past my peak, even tho it
was a much lower peak than i should have been reaching. Creating a
sentient Ai tells me the machine is valued over me.
I just got a price for *me* to get dsl. It will be installed 60 miles from me,
cost $80/month (plus installation), in a section south of Birmingham noted for
high crime and drug abuse rates. There will be no security for the installation,
and it's a 2hour commute to drive there. Either that, or i learn Perl to get the
nix shell in Netherlands operational.
Kat
8. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 06, 2002
|Then you're simply making an expert system that can play tic-tac-toe.
|
|And playing tic-tac-toe is no easy feat. You say that's the basics,
but
|you've gone waaaaay beyond the basics to suggest that our AI entity
can play
|a game and learn from it!
|
|I'm thinking just a virtual worm (of our own design) in a virtual
bowl of
|dirt where we sprinkle some water and food every once-in-a-while and
see
|what happens.
|
|To be truly AI, you have to think "primitive." Nay, "very primitive."
:)
I think you missed how I meant it.
expert system would be one one of 'implementing' it, but that isn't
how or what I mean. I had a very long discussion with some research
groups here at uni awhile ago, and we generally agreed that starting
with a basic brain and teach it to play tic-tac-toe was one of the
best starting points for a little AI project.
Firstly. Any AI will of course have to come with some pre-programmed
specifics otherwise they would just sit and not interact at all. So
say it starts of 'know' how to place pieces on the board, and that it
is taking in turns.
Then you need it to understand that if you say, "You lost" that it is
a negative thing, and "You Win" is a good thing.
That way we could have it 'learn' which is the best combinations to
play give how the other player plays. Tic-tac-toe is I believe on of
the easiest games to start with. Easy to see how it is progressing.
Then maybe would could move onto the 3d version of tic-tac-toe, that
would give a bigger chance for more complex AI techniques to be
tested, with pattern match, since there are more options.
etc, etc.
9. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 06, 2002
|I'd like to see some information, tutorials, or libraries that can
use
|simple neural nets. That's the sort of thing I want to get involved
in, but
|I don't know much about them, or how they are implemented.
Open google and run a search for neural net. There a some very good
gui neural net programs out there that let you build a net visually
and teach it 'visually' etc. Ther even come with some sample nets
premade. One I downloaded even a ocr net made that worked pretty well.
I can't remember there names thou. If you can't find any good ones,
tell me, and I'll go look for you.
Cheers,
Dan
10. RE: The A.I. Project II
Perhaps I should clarify...
I'd like to see a EUPHORIA neural net library.
=====================================================
.______<-------------------\__
/ _____<--------------------__|===
||_ <-------------------/
\__| Mr Trick
>From: dm31 at uow.edu.au
>Subject: RE: The A.I. Project II
>Date: Thu, 07 Nov 2002 14:49:45 +1100
>
>
>|I'd like to see some information, tutorials, or libraries that can
>use
>|simple neural nets. That's the sort of thing I want to get involved
>in, but
>|I don't know much about them, or how they are implemented.
>
>Open google and run a search for neural net. There a some very good
>gui neural net programs out there that let you build a net visually
>and teach it 'visually' etc. Ther even come with some sample nets
>premade. One I downloaded even a ocr net made that worked pretty well.
>
>I can't remember there names thou. If you can't find any good ones,
>tell me, and I'll go look for you.
>
>Cheers,
>Dan
>
>==^^===============================================================
>This email was sent to: mistertrik at hotmail.com
>
>
11. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 06, 2002
|Perhaps I should clarify...
|
|I'd like to see a EUPHORIA neural net library.
Check the archives. I'm sure there is one an old n-net lib in there.
12. RE: The A.I. Project II
C.K.,
The environment itself brings a fascinating question.The idea of an a.i
existing within the computers architecture is very appealing to me.But
the question is, how will we see its progress? Or if it can express
itself,how will we see what it "says"? The computer is all about
input/output right? The logical "output" would initially be text,English
language text! Now lets take it to the next level: lets
say we program a strong instinct within it to live, a strong,gut
instinct to survive,and lets say it needs to eat to survive,so now
we have a being that is hungry,very hungry.What does a computer "eat"?
It eats input!We feed it information!If its online, the world feeds
it information!
On to digestion, what is digestion to us? We take what our bodies
need and send the waste to the draught.What is digestion to an a.i?
When the a.i digests,it will be processing the input,but beyond the
basic computers processing cycle,the a.i. is smarter,it filters the
information,it finds relevance and it stores the information in its
hardware,its body.What it does not find relevance for in its body
of information,it discards.
On to output,now that the a.i. has information, I guess we could
ask it questions and it would gives us answers,but all that is is
a glorified database...
This is where the intelligence part of artificial intelligence comes
in.The a.i. only stores the information that it WANTS to store,based
on the REASONS it has found RELEVANCE TO ITS OWN GOALS.We establish
these goals,we create its instincts.
Furthermore as to the question of the virtual fishtank and the virtual
worm.Those would not be emulations if the a.i. chose to
represent himself as such.Graphics are just another way for the
computer to output information.While we know the a.i. really exists
in electronic pulses,we could be able to see it in this parallel
graphical world.For instance,after we type in input,we could see a
peice of food being dropped into its cage,or its fishtank whatever
the case may be.What if it learned to create its own graphics and
chose to represent itself in a different way,graphically?
So I would choose to have both environments,first a text based
environment,and eventually a parallel,graphical environment.Then
I would move towards finding different methods of input such as
video cameras and audio devices.
What do you think?
JDUBE
>From: "C. K. Lester" <cklester at yahoo.com>
>Reply-To: EUforum at topica.com
>To: EUforum <EUforum at topica.com>
>Subject: RE: The A.I. Project II
>Date: Wed, 6 Nov 2002 21:07:11 -0600
>
>
> > As far as finding out what is instinctual for a worm,first of all
> > that would entail finding out what kind of worm we are talking about,
> > and then doing research on that particular kind.
>
>I think it'd be easier to invent our own virtual species of worm that lives
>on a PC.
>
>The problem is, we want to create a sentient entity, right?
>
> > And thats not to say
> > we couldn't just invent a type of worm that has instincts that we
> > provide.After all, we will become,in effect,its creator!
>
>Probably the easiest option, wouldn't you say?
>
>There are two approaches:
>
>1. create an actual AI entity whose "world" is the PC (+internet). The PC's
>hardware thus becomes its "body."
>2. EMULATE an AI entity in an artificial (emulated) virtual environment (a
>worm in a glass bowl, for instance).
>
>I don't know which would be easier. I'd like to hear thoughts on it before
>I
>delve into it more.
>
>
>
13. RE: The A.I. Project II
Kat wrote:
>Either that, or i learn Perl to get the
> nix shell in Netherlands operational.
Kat,
Maybe I shouldn’t write this on the list ... ( to late )
If your "serious" about writing commercial software (whatever type)
you should seriously consider what other languages are
available and the pros and cons of each.
I like Euphoria, and it's lots of fun. I particularly enjoy
the mailing list and have to admit the Euphoria community
is the best bunch of people I have come across on the net.
Saying that, I would be very careful choosing to write commercial
software with Euphoria. The lack of threading and exceptions are
the two biggest drawbacks for any “server” type of software.
The lack of good quality libraries in numerous areas of computing
and the lack of people using the current crop of libraries must be
considered before jumping into a commercial product with Euphoria
aswell.
I haven’t really followed the mailing list threads up to your last
email so I’m not really sure what your trying to do … and I’m not
saying don’t use Euphoria ... but I am saying you should look at the
alternatives :)
Maybe you should document what features you find critical (in a
language) For you to complete your project and post it to the list
and see what suggestions come back.
I’d be happy to discuss details with you further if you want.
Regards,
Ray Smith
http://rays-web.com
14. RE: The A.I. Project II
On 6 Nov 2002, at 23:15, dubetyrant at hotmail.com wrote:
>
> C.K.,
> The environment itself brings a fascinating question.The idea of an a.i
> existing within the computers architecture is very appealing to me.But
> the question is, how will we see its progress? Or if it can express
> itself,how will we see what it "says"? The computer is all about
> input/output right? The logical "output" would initially be text,English
> language text!
Or not English! How about Italian or Malaysian!
> Now lets take it to the next level: lets
> say we program a strong instinct within it to live, a strong,gut
> instinct to survive,and lets say it needs to eat to survive,so now
> we have a being that is hungry,very hungry.What does a computer "eat"?
> It eats input!We feed it information!If its online, the world feeds
> it information!
Well, if you have enough bandwidth, anyhow <sigh>.
> On to digestion, what is digestion to us? We take what our bodies
> need and send the waste to the draught.What is digestion to an a.i?
> When the a.i digests,it will be processing the input,but beyond the
> basic computers processing cycle,the a.i. is smarter,it filters the
> information,it finds relevance and it stores the information in its
> hardware,its body.What it does not find relevance for in its body
> of information,it discards.
> On to output,now that the a.i. has information, I guess we could
> ask it questions and it would gives us answers,but all that is is
> a glorified database...
<tongue in cheek>
Tell me about it.
</tongue in cheek>
> This is where the intelligence part of artificial intelligence comes
> in.The a.i. only stores the information that it WANTS to store,based
> on the REASONS it has found RELEVANCE TO ITS OWN GOALS.We establish
> these goals,we create its instincts.
> Furthermore as to the question of the virtual fishtank and the virtual
> worm.Those would not be emulations if the a.i. chose to
> represent himself
Or herself. Remember, on Star Trek, Data's offspring "chose" to be female.
> as such.Graphics are just another way for the
> computer to output information.While we know the a.i. really exists
> in electronic pulses,
And mammals (humans included) exist as electrochemical impulses.
> we could be able to see it in this parallel
> graphical world.For instance,after we type in input,we could see a
> peice of food being dropped into its cage,or its fishtank whatever
> the case may be.What if it learned to create its own graphics and
> chose to represent itself in a different way,graphically?
> So I would choose to have both environments,first a text based
> environment,and eventually a parallel,graphical environment.Then
> I would move towards finding different methods of input such as
> video cameras and audio devices.
> What do you think?
I asked someone about using some program to make 3D slices of a figure for
a bot output on a screen. Naturally, this whole project fell thru also. It's
been
done, but not by anyone i know. Besides, to do this online, big files need to
be sent, meaning bandwidth again. So i'll be sitting here watching everyone
else do it, as usual.
Kat
15. RE: The A.I. Project II
On 6 Nov 2002, at 21:33, C. K. Lester wrote:
>
> > Maybe we should start with the very basics first.
> >
> > Maybe get some 'brain base' to be able to play tic-tac-toe...
>
> Then you're simply making an expert system that can play tic-tac-toe.
>
> And playing tic-tac-toe is no easy feat. You say that's the basics, but
> you've gone waaaaay beyond the basics to suggest that our AI entity can play a
> game and learn from it!
Tic-tac-toe is a very restricted domain, and a very restricted solution set. You
can make a "computer" that plays that game with relays. I worked out the
logic for it in the 7th grade. It didn't learn.
Kat
16. RE: The A.I. Project II
On 7 Nov 2002, at 14:29, mistertrik at hotmail.com wrote:
>
> I'd like to see some information, tutorials, or libraries that can use
> simple neural nets. That's the sort of thing I want to get involved in, but I
> don't know much about them, or how they are implemented.
Think "analog neurons". Think "cells" with multiple inputs, multiple outputs,
and mediated by the "chemicals" in the bloodstream and intracellular space.
Most neural nets have a few layers of cells with some feedback mechanism,
some are moderated with a traditional computer, by changing the thresholds
at which what makes what cells fire when. This can be done easily with Eu
sequences.
Kat
17. RE: The A.I. Project II
On 7 Nov 2002, at 6:08, Ray Smith wrote:
>
>
> Kat wrote:
>
> >Either that, or i learn Perl to get the
> > nix shell in Netherlands operational.
>
> Kat,
>
> Maybe I shouldn’t write this on the list ... ( to late )
>
> If your "serious" about writing commercial software (whatever type)
> you should seriously consider what other languages are
> available and the pros and cons of each.
If i had the exu socks code for bi-directional http relaying on a freebsd shell,
i
can probably get the rest of the necessities done overnight in Eu. If it's done
in Perl, i haveto get someone else to write the *entire* program, and i have a
learning curve before i can change anything. Both perl 5.8.0 and the latest
exu are installed on the box. I have permission to run the proxy within limits,
and if i pass those limits my connection will be swamped anyhow and i hope
to have income by that time, if that time comes, and i can blow the money
on bandwidth more locally to me, so i won't need the shell. The shell will
provide bandwidth and a static ip to me.
Kat
18. RE: The A.I. Project II
dm31 at uow.edu.au wrote:
> I had a very long discussion with some research
> groups here at uni awhile ago, and we generally agreed that starting
> with a basic brain and teach it to play tic-tac-toe was one of the
> best starting points for a little AI project.
Therein lies your problem... Where ya gonna get that "basic brain?"
Are you going to fill that brain with knowledge, or are you going to
fill it with the tools to gain knowledge (or both)?
> So say it starts off 'knowing' how to place pieces on the board,
> and that it is taken in turns.
Even before that you have to start with it knowing what a piece is, what
a board is, what a game is, what competition means, what winning and
losing means, what a turn is, what an opponent is, etc. etc...
Give it "tic-tac-toe" instincts and let it live and learn. (The problem
with any AI entity is going to be programming the instincts.)
> Then you need it to understand that if you say, "You lost" that
> it is a negative thing, and "You Win" is a good thing.
At this point it needs to be understanding "consequences" and its
relation to the "win/lose" instinct (or something like that). You say
it's a negative thing, but has it learned what a "negative" thing is?
How is it going to care what is negative/positive? You might say, "You
lost" and it might respond, "So what?" or "What does that mean?"
> That way we could have it 'learn' which is the best combinations to
> play give how the other player plays. Tic-tac-toe is I believe on of
> the easiest games to start with. Easy to see how it is progressing.
You've got to start off with an "infant TTT-playing AI entity" that
functions on instinct alone, then let it "grow," interacting with it and
"raising" it, or guiding it. Otherwise, you're going to have to do more
programming at the start than is humanly possible.
19. RE: The A.I. Project II
dubetyrant at hotmail.com wrote:
> The environment itself brings a fascinating question.The idea of an a.i
> existing within the computers architecture is very appealing to me.
Yeah, I like the thought, too. This kind of entity wouldn't need to eat
or sleep or whatever. It's the kind that should develop extremely fast
because there are limited hurdles on its path to full-potential
development.
> But the question is, how will we see its progress?
Let it act (and interact) in a window. It's not going to have any body
in which to convey any language, so it must be totally verbal... UNLESS
you want to provide a graphical body for it (on screen)... but that
would be more complexity than we need right now.
> language text! Now lets take it to the next level: lets
> say we program a strong instinct within it to live, a strong,gut
> instinct to survive,and lets say it needs to eat to survive...
If you have a purely code-based entity, it won't need to eat. And it
will have no need for sleep. It will also have absolutely no way to
prevent its creator from "turning it off" (killing it), which might be a
scary proposition for it. :)
Regardless, no eating/digesting required.
> So I would choose to have both environments, first a text based
> environment,and eventually a parallel, graphical environment.
Yes.
Its learning to understand text input would be analogous to a human
baby's learning language... At first it wouldn't understand what we were
saying... But, eventually, it would notice patterns and maybe start to
repeat those patterns and associate those patterns with environmental
objects ("dada" = that thing that induces a laugh response by touching
me, "mama" = that thing that satisfies my hunger response and takes away
my "uncomfortable-poopy-diaper" response). "No" would be associated with
things in the negative. "Yes" in the opposite.
Of course, we're not talking about a worm here. Maybe we should just
talk about a worm...? I dunno.
20. RE: The A.I. Project II
jslezak at luc.edu wrote:
> How about a dog AI.
I think we need to stay away from emulating an existing creature, unless
it is very simple. When you suggest using a dog, now we really have to
get into physiology... the brain signaling muscles to move.
Let's just create our own worm.
21. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 07, 2002
Before you read my replies below, plz note that I am not writing this
as a suggest for true AI. For us to make true AI while doing this is
not quite same :P There are many ppl who have dedicated there lives to
this thing. Done PhD's on it, and they haven't succeed with true AI
yet. Maybe we should start of with a more realistic goal of a proggie
that can 'learn' how to do something better as it collects more data.
Go read some of the indepth AI papers out there to get an idea of
where ppl are heading with AI.
Check out some papers on Neural nets too. They are alot more complex
then may appear at first glance. try here or interesting and
informative reading:
http://www.emsl.pnl.gov:2080/proj/neuron/neural/what.html
My idea is just something to start with, get something with result
working, something that can be built on.
|> I had a very long discussion with some research
|> groups here at uni awhile ago, and we generally agreed that
starting
|> with a basic brain and teach it to play tic-tac-toe was one of the
|> best starting points for a little AI project.
|
|Therein lies your problem... Where ya gonna get that "basic brain?"
errr. Thats what your programming :P
|Are you going to fill that brain with knowledge, or are you going to
|fill it with the tools to gain knowledge (or both)?
|
|> So say it starts off 'knowing' how to place pieces on the board,
|> and that it is taken in turns.
|
|Even before that you have to start with it knowing what a piece is,
what
|a board is, what a game is, what competition means, what winning and
|losing means, what a turn is, what an opponent is, etc. etc...
|Give it "tic-tac-toe" instincts and let it live and learn. (The
problem
|with any AI entity is going to be programming the instincts.)
Not really. What does it care at this point what a 'piece' is, or what
competition is. It just needs to be able to distinguish between a
success and a fail. Start with the basics, and work your way up.
|> Then you need it to understand that if you say, "You lost" that
|> it is a negative thing, and "You Win" is a good thing.
|
|At this point it needs to be understanding "consequences" and its
|relation to the "win/lose" instinct (or something like that). You say
|it's a negative thing, but has it learned what a "negative" thing is?
|How is it going to care what is negative/positive? You might say,
"You
|lost" and it might respond, "So what?" or "What does that mean?"
read above.
|> That way we could have it 'learn' which is the best combinations to
|> play give how the other player plays. Tic-tac-toe is I believe on
of
|> the easiest games to start with. Easy to see how it is progressing.
|
|You've got to start off with an "infant TTT-playing AI entity" that
|functions on instinct alone, then let it "grow," interacting with it
and
|"raising" it, or guiding it.
Hmmm. Isn't that what I said??
|Otherwise, you're going to have to do more
|programming at the start than is humanly possible.
With something as simple TTT, you could program a computer TTT that
could never lose a series if statements in a short period of time. :P
There isn't that many combinations on a 3x3 board....
Hope that cleared it up a bit.
Dan
22. RE: The A.I. Project II
dm31 at uow.edu.au wrote:
> Before you read my replies below, plz note that I am not writing
> this as a suggestion for true AI.
Oh. Well, when people say "AI" I expect they mean "AI." If you're
talking about an expert system (ES), then that's not AI. Doesn't even
come close to resembling AI.
At least that clears a few things up... ;)
> Done PhD's on it, and they haven't succeed with true AI yet.
That's 'cuz they're dumb. Okay, just kiddin'. They are just approaching
it from the wrong angle, that's all. And we don't have the technology
yet.
> Maybe we should start of with a more realistic goal of a
> proggie that can 'learn' how to do something better as it
> collects more data.
Aren't there programs that already do this? What are you going to
contribute to that particular topic? Is it just to do it in Eu?
> Go read some of the indepth AI papers out there to get an idea of
> where ppl are heading with AI.
I'm going to guess that the general state of the industry is that
they're heading no where. Too many blocks, in hardware, in software, and
philosophy.
> My idea is just something to start with, get something with result
> working, something that can be built on.
You can't "build on" an expert system. You haven't given it the hardware
or software to be able to be AI. You can give an AIE expert system
knowledge, but it won't go the other way around. You can't make an ES
sentient.
> |Therein lies your problem... Where ya gonna get that "basic brain?"
>
> errr. Thats what your programming :P
You suggest that we are going to emulate a brain via software... I
propose that you better find a very primitive brain to emulate.
> |Even before that you have to start with it knowing what a piece is,
> what
> |a board is, what a game is, what competition means, what winning and
> |losing means, what a turn is, what an opponent is, etc. etc...
> |Give it "tic-tac-toe" instincts and let it live and learn. (The
> problem
> |with any AI entity is going to be programming the instincts.)
>
> Not really. What does it care at this point what a 'piece' is, or
> what competition is. It just needs to be able to distinguish
> between a success and a fail. Start with the basics, and work your
> way up.
True, an expert system doesn't need to "know" all these things, but an
AI entity should. Since you're talking ES and not AI, what you say is
valid.
And starting this way will NOT let you get to AI... just a more
efficient expert system. You can't start with ES and get to AI.
> |You've got to start off with an "infant TTT-playing AI entity"
> |that functions on instinct alone, then let it "grow,"
> |interacting with it and
> |"raising" it, or guiding it.
>
> Hmmm. Isn't that what I said??
No. Not even close. You're proposing an expert system, not an AI entity.
> With something as simple TTT, you could program a computer TTT that
> could never lose a series if statements in a short period of
> time. :P There isn't that many combinations on a 3x3 board....
There was the suggestion of Connect-4. Sounds reasonable to me.
23. RE: The A.I. Project II
> http://www.emsl.pnl.gov:2080/proj/neuron/neural/what.html
Site: "Another way of classifying ANN types is by their method of
learning (or training), as some ANNs employ supervised training while
others are referred to as unsupervised or self-organizing."
Me: "unsupervised or self-organizing" - obviously they don't know what
they're talking about. No intelligent being that we know of is able to
go from instinct to intelligence on its own, simply because it must come
from an instinct-only stage, during which it is at its most vulnerable.
Site: "Supervised training is analogous to a student guided by an
instructor."
Me: The presupposition here is that the student has LEARNED HOW TO
LEARN. With AI, though, you cannot just start out with an advanced
learner. It's IMPOSSIBLE. An "advanced learner" is way too complex for
you to create in the first place. When you try and create an "advanced
learner," you've simply created an expert system with all your own
biases. The AI entity has to start out much simpler, with the tools
(hardware+instincts) to learn (just like the only intelligent creatures
we know must do).
Site: "Unsupervised algorithms essentially perform clustering of the
data into similar groups based on the measured attributes or features
serving as inputs to the algorithms. This is analogous to a student who
derives the lesson totally on his or her own."
Me: What beginning intelligent creature have you ever known was able to
learn on its own from the start? Sure, LATER, after it became an
advanced learner did it learn how to acquire knowledge itself... but it
couldn't start like that.
24. RE: The A.I. Project II
On 7 Nov 2002, at 18:43, C. K. Lester wrote:
>
> > http://www.emsl.pnl.gov:2080/proj/neuron/neural/what.html
>
> Site: "Another way of classifying ANN types is by their method of
> learning (or training), as some ANNs employ supervised training while
> others are referred to as unsupervised or self-organizing."
>
> Me: "unsupervised or self-organizing" - obviously they don't know what
> they're talking about. No intelligent being that we know of is able to
> go from instinct to intelligence on its own, simply because it must come
> from an instinct-only stage, during which it is at its most vulnerable.
It works in reverse too.
> Site: "Supervised training is analogous to a student guided by an
> instructor."
>
> Me: The presupposition here is that the student has LEARNED HOW TO
> LEARN. With AI, though, you cannot just start out with an advanced
> learner. It's IMPOSSIBLE. An "advanced learner" is way too complex for
> you to create in the first place. When you try and create an "advanced
> learner," you've simply created an expert system with all your own
> biases. The AI entity has to start out much simpler, with the tools
> (hardware+instincts) to learn (just like the only intelligent creatures
> we know must do).
I disagree. The parts you don't kow are the quantity and degree of weightings
of incentives and disincentives, and those unknowns in knowledge which
may become guesses in action. The rest is pattern matching:
"my stomach hurts"
-- has this happened before?
-- match this to incentives pattern map
-- match up a pattern of actions to what we are in now
"want to cure this, need food"
-- match "food" to the actual item irl
-- match "food" to what's in the environment
-- match a pattern of movement to ingest the food
-- match these patterns to patterns of outcomes
-- (do not walk up to live elephant and begin eating)
-- expect pattern to continue post-ingestion
etc
> Site: "Unsupervised algorithms essentially perform clustering of the
> data into similar groups based on the measured attributes or features
> serving as inputs to the algorithms. This is analogous to a student who
> derives the lesson totally on his or her own."
That's a pattern matching or modification. Truely unexpected pattern
matching leads people to run naked thru the streets shouting "Eureka!" if it's
good, and become snipers if it's bad.
> Me: What beginning intelligent creature have you ever known was able to
> learn on its own from the start? Sure, LATER, after it became an
> advanced learner did it learn how to acquire knowledge itself... but it
> couldn't start like that.
Only if a learning activity became a function of enabling an instinct, resulting
in positive feedback. I have discovered thru personal experience and machine
programming that if there is no feedback, there is no such function
generated, no exercizing of memory expansion, no change in incentive
weighting, etc. Hence you cannot have the Ai generate it's own node/neuron
weighting. Even if it did get some learning, you'll never be able to predict
what the outcome is.
Given one person on a deserted island, with a pocket knife, infected with
AIDS, what incentive do they have to build an electron microscope to
research the mechanism of antibodies on HIV, and develop a cure and live
happily ever after? The Ai hasto check the weighting. If it has OCD and is
really bored, it might try to find a way to generate electricity and build the
EM. Otherwise it might just try to get comfy and *seemingly* violate self
preservation instincts with euthanasia. We can't predetermine this in our
hardcoding, else it won't be an Ai. If you hardcode or build hardware that
stresses *your* values, you have a slave.
Kat
25. RE: The A.I. Project II
Yes, I did that also.
I chose to remove all the code for if it lost a game, because it never did.
Every game it'd either win or draw.
Kinda boring to play against though.
=====================================================
.______<-------------------\__
/ _____<--------------------__|===
||_ <-------------------/
\__| Mr Trick
>From: Kat <kat at kogeijin.com>
>Subject: RE: The A.I. Project II
>
>
>On 6 Nov 2002, at 21:33, C. K. Lester wrote:
>
> >
> > > Maybe we should start with the very basics first.
> > >
> > > Maybe get some 'brain base' to be able to play tic-tac-toe...
> >
> > Then you're simply making an expert system that can play tic-tac-toe.
> >
> > And playing tic-tac-toe is no easy feat. You say that's the basics, but
> > you've gone waaaaay beyond the basics to suggest that our AI entity can
>play a
> > game and learn from it!
>
>Tic-tac-toe is a very restricted domain, and a very restricted solution
>set. You
>can make a "computer" that plays that game with relays. I worked out the
>logic for it in the 7th grade. It didn't learn.
>
>Kat
>
>
>
26. RE: The A.I. Project II
- Posted by dm31 at uow.edu.au
Nov 07, 2002
|Oh. Well, when people say "AI" I expect they mean "AI." If you're
|talking about an expert system (ES), then that's not AI. Doesn't even
|come close to resembling AI.
No sorry. An ES system is complete;y pre-programmed, it does not
'learn' once it is setup. It starts of with a large collection of
rules. What I suggested does NOT.
|That's 'cuz they're dumb. Okay, just kiddin'. They are just
approaching
|it from the wrong angle, that's all. And we don't have the technology
|yet.
Well. There are some really advanced systems out there that are really
exciting and show alot of promise. I was once pretty wrapped up in
this field and tried to keep current. Some of the stuff they have done
with "Robotic" personalilities that were learning english, picking up
slang, and identifying via video input was quite suprising
|> Maybe we should start of with a more realistic goal of a
|> proggie that can 'learn' how to do something better as it
|> collects more data.
|
|Aren't there programs that already do this? What are you going to
|contribute to that particular topic? Is it just to do it in Eu?
Once I have finished my final exams at uni and get some time, I mostly
likely will start to write something along these lines in EU.
|> Go read some of the indepth AI papers out there to get an idea of
|> where ppl are heading with AI.
|
|I'm going to guess that the general state of the industry is that
|they're heading no where. Too many blocks, in hardware, in software,
and
|philosophy.
read above ^^
|> My idea is just something to start with, get something with result
|> working, something that can be built on.
|
|You can't "build on" an expert system. You haven't given it the
hardware
|or software to be able to be AI. You can give an AIE expert system
|knowledge, but it won't go the other way around. You can't make an ES
|sentient.
Read what I said at the start. There is a big difference between a ES
and what I proposed.
|> |Therein lies your problem... Where ya gonna get that "basic
brain?"
|>
|> errr. Thats what your programming :P
|
|You suggest that we are going to emulate a brain via software... I
|propose that you better find a very primitive brain to emulate.
It was in quotes for a reason. "Brain" was refering to the section of
software that did the learning etc. IE, the neural net, pattern
matching code, etc, etc, excluding interface & wrapper. It is commonly
used to refer to this when discussing *AI*
|> |Even before that you have to start with it knowing what a piece
is,
|> what
|> |a board is, what a game is, what competition means, what winning
and
|> |losing means, what a turn is, what an opponent is, etc. etc...
|> |Give it "tic-tac-toe" instincts and let it live and learn. (The
|> problem
|> |with any AI entity is going to be programming the instincts.)
|>
|> Not really. What does it care at this point what a 'piece' is, or
|> what competition is. It just needs to be able to distinguish
|> between a success and a fail. Start with the basics, and work your
|> way up.
|
|True, an expert system doesn't need to "know" all these things, but
an
|AI entity should. Since you're talking ES and not AI, what you say is
|valid.
AS I said, it isn't an ES system. A ES system is a static database of
rules and results. An ES system uses input data to do through and test
rules. It then checks which result is most probable depending on which
rules are valid. And ES system will repeatedly give you the same
result given the same input. What I suggested creates & changes it's
own rules as it is guided along it's learning curve, just like we do.
We we were a very young baby, did we know what "Food" was. No. It was
just instinct to eat stuff that was put near our mouths, hence a baby
must be watched for eatting small toys etc. We 'learn' what food is as
we grow up a bit. Just think of this as a half *AI* at a very young
stage.
|And starting this way will NOT let you get to AI... just a more
|efficient expert system. You can't start with ES and get to AI.
Above ^^ This is very different from an ES, and infact forms a basis
for cores of an attempted to do AI.
|> |You've got to start off with an "infant TTT-playing AI entity"
|> |that functions on instinct alone, then let it "grow,"
|> |interacting with it and
|> |"raising" it, or guiding it.
|>
|> Hmmm. Isn't that what I said??
|
|No. Not even close. You're proposing an expert system, not an AI
entity.
Yes it was. I said give it the basic abilities to play TTT, with out
telling 'how'. Then let it "learn" (grow) by playing againsted it a
giving it feed back as to a success or a fail. Which is what you say
above.
|> With something as simple TTT, you could program a computer TTT that
|> could never lose a series if statements in a short period of
|> time. :P There isn't that many combinations on a 3x3 board....
|
|There was the suggestion of Connect-4. Sounds reasonable to me.
Yes Connect-4 would be another easy game to start with.
