Re: A request for information please.

N wrote:
On 15 May, 23:44, Stephen Harris <cyberguard-1...@xxxxxxxxx> wrote:
The underlying goal of AI is to write a program that will appear
or behave as if it has human level intelligence. The best test
for this is called the Turing Test (TT). A program passes the TT
if it can have an email conversation with you for an extended
period of time and convince you (without you knowing beforehand)
that it is human, because it thinks fairly rationally like a human.
This is a lesser goal than having a walking about human level robot.

The program can have several languages for expressing itself. The
hard part is to get it to seemingly understand what you say so that
it can respond with appropriate, rather than odd, replies. To carry
on a conversation remembering and relating to earlier parts of the
conversation. Learning from examples would be important because a
human can learn from example. Analogies are a major part of this.

So the goal is to create an AI program and this means writing the
program in a programming language. There are sub-goals, such as
automatic layout of a web page, so such "expert" programs are useful
but still need to be written in some formal programming language.
There has been almost total failure in 50 years for TT passing programs.

Design wise, I've read from others that the influence of modern
'instant' layouts has given a bland and impersonal quality to
presentation, one to one design practice on a project means that
the designer can be absolutely precise about what is required,
taste, aesthetics, culture, what sort of people the media design
is aimed for, readership etc all need to be considered. Layouts
need to be regularly updated and I imagine it to be the same
working on a language base, where there are phrases and oblique
references and where content varies from time to time.


Yes, they would like to build a program that can read a paper and
then provide a summary. Or a Google translation program that can
do better than literal word for word substitution which makes some
very awkward translations of some phrases.

Some programming languages are more visually oriented than others, so
that a background in graphics design might point to using a visual
approach in trying to create an AI program (if that is your goal).

Wolfram thinks the whole universe might have evolved from a simple
cellular automata (CA) rule to all the current complexity. Since
there are so many CA rules it is hard to discover the rule which
will unfold into a desired outcome. The idea of emerging a mind is
part of Artificial Life, a related field to to AI. Shape Grammars:

I read the overview of 'A New Kind of Science' on MIT,

so, a single cellular automata, with a set of constraints and its
evolution bounded by the limits of its own design, etc. but then
that does'nt include 'contingencies' I mean it would have to
develop into something someplace wouldn't it? the external 'somewhere'
affecting it's A.IQ. as I havn't read the book I'm unsure if he
means he's 'seeded' a hypothetical space to develop an enviroment
in which to 'grow' or evolve the CA's


I have only played with CAs a little bit. The Game of Life was
invented by Conway about 40 years ago. It is a cellular automata
with interesting properties that have fascinated lots of people.

http://mathworld.wolfram.com/Life.html
"Amazingly, Life is a universal cellular automaton, in the sense that
it is effectively capable of emulating any cellular automaton, Turing machine, or any other system that can be translated into a system known to be universal. The outlines of a proof for life's universality were given by Berlekamp, Conway, and Guy (1982) and independently by Gosper (Gardner 1983, pp. 250-253). Around 2000, a Turing machine that can be extended to a universal Turing machine was explicitly implemented in life by P. Rendell (Rendell, Adamatzky 2001). While Rendell's machine can be made into a "true" universal computer simply by making his tape infinite, he neither noted this fact nor provided an actual construction of a universal Turing machine. Subsequently, on November 11, 2002, P. Chapman constructed a life pattern based on D. Hickerson's "sliding block memory" approach that implements the actions of a universal register machine. Unlike the finite tape of Rendell's Turing machine, the values in the registers of Chapman's machine are unbounded, making it a true model of universal computation in the game of life."

SH: I think that there are a countable infinity of ordinary Turing
Machines. Then there are Universal Turing Machines which can simulate
any of the particular ordinary Turing Machines (TMs). The quote above
says that the cellular automata, named Life, is capable of universal
computation of Turing computable functions, so there is an equivalent
computing capability for solving problems between UTMs, universal cellular automata and computers. If I didn't say that quite correctly
somebody will probably correct me. TMs are not exactly physically
realizable because they have a potentially infinite tape that serves
in place of memory for a PC; thus PCs have a finite amount of memory.

There is a huge number of possible CAs. Not all of them are capable of
universal computation. Now, if it turns out that the human mind, or all the functions of human intelligence, are Turing-computable functions,
then it may well be possible to create such a program that will run on
a computer. Then there will also be some universal cellular automaton rule, or concatenation of CA rules (shape grammar) which instantiates the same properties as the TM approach can instantiate. I'm not sure
that I've read this or I've figured it out, so this idea needs to be
confirmed.


"The research objective herein is to exploit the potential of CA by
presenting an approach for using shape grammar [5] (a formal set of
rules applied to shapes to generate a language of design; allows the
visualization of the desired form and function of the rules)

But does this mean CA's develop into framework structure, like
neurons and consequently an environment in which consciousness
and cognitive events only occur, & is he saying that ALL structures
of expression, including the development of grammars etc, can be
founded upon that?.

Computable structures, and it is an open question about whether all
physical structures can be simulated by Turing-function approximation.
Wolfram thinks the entire complexity of the universe could have evolved
out of some simple CA rule unfolding though time; he thinks the universe
is a gigantic cellular automaton. But not a lot of people think so.

Since CAs became popular there are many CA programs and adaptions
of the Game of Life, that will run on your computer. You can find
web sites that run CA rule simulations so that you can see how the
rules evolves from a simple beginning to more complex strutures.
Life has several interesting structures evolve, I would say virtually.
comp.theory.cell-automata is a newsgroup with lots of discussion,
Tim Tyler is very knowledgeable. I think they have a FAQ.

I don't know, considering the number of neurons/computations we
make, by the time we're aware of something, we're still fairly
limited by some sort of perceptional uptake. I think spoken
language and communications must have developed a heck of lot
late on in our evolution tho.


Yes, so they say about language. I've read that our senses receive
2 million (or billion?) bits per second, and that some evolutionary
filtering process then only passes on 2,000 bits per second from the
unconscious to our conscious awareness.

to derive
CA rules. The essential scope of this investigation falls within the
domain of self-generative system architecture, with the specific
methodology entailing the use of a shape grammar for establishing the
rules of the design process (at both the elemental and organizational
levels), followed by a CA approach for actually generating the
creative design space. The shape grammar thus expresses the physics
of the formfunction and becomes transcribed into CA rules and their
conditional neighborhoods, with the CA rules generating the design space.

the creative design space, or conditional neibourhoods, influences
the the CA rules and physics in some way to feedback into the
creative design space?


Yes, as Wolfram would say, very simple beginnings produce very complex
patterns. I've read that people get addicted to watching the Game of Life. http://x.i-dat.org/~csem/UNESCO/8/index.html

The use of shape grammars affords to the system architect the capability
for assuring that design proceeds by rules that embody the relevant
principles of physics and can select design candidates that meet
stability, robustness, aesthetics, cost and other requirements, thereby
managing an otherwise possibly explosive design space. At the same time,
shape grammars allow the system architect to explore a diverse variety
of design styles, providing opportunities for the emergence of
potentially useful higher order components and modular structures. The
system architect's role is thus to create a design space of
conceptualizations and select good system architectures from this design
space for a specification, to determine the design physics and selection
rules to be implemented, to develop the shape grammars to reflect these
rules at the modular and hierarchical levels, and to program the CA to
capture these rules and output a design catalog of the best candidates
that meet the specification."

I'd be sure to say 'conceptualization' only if I were absolutely
sure 'it' imitated natural processes of thought, and if the
specification is intelligences, suitability to perform a miracle
in its environment.


This is a murky area IMO. There is a Grand Master level chess program
that can beat almost any human. It outputs moves that some good players
would make and some beyond their ken. It does this by brute force
calculation of millions of possible moves which a human can't do. So
I don't think it imitates how humans play chess, although it produces
many of the same moves Masters would make, it isn't the same method.

"Formal grammars are systems of rules for characterizing the structure,
or the syntax, of sentences in natural and artificial languages. Shape
grammars are a geometrical design adaptation of Noam Chomsky's formal
(phrase structure or transformational) grammars [19] and are recursively
enumerable, having the capability of producing unrestricted languages
[20]. Thus, shape grammars are systems of rules for characterizing the
composition of designs in spatial languages."

nice idea to think of language as 'modules' in 3D space I suppose,
(yep I know its a reversal ;) although there's so many ways to link
systems together,hmmm if I had more time I wouldn't half fancy
exploring that


Did you notice there is a $25,000 prize for solving a cellular automata
theory question? Now you have incentive. I doubt that it will be easy.

SH: Scheme (free) and Lisp are languages often used to write AI
programs and they are more symbolic than the CA/shape grammar approach.

Even just in conversation, the environment of possible discussions is
huge and the problem for AI is to select the appropriate responses to
match the ever changing and revised contexts of the environment. A
healthy adult has much the same physical equipment due to genes, so
we all see a dog (Nature), sense time passing, expect cause and effect,
etc; but then each mind gets its own unique set of shaping experiences
(Nurture) causing differences of opinion, morals, and language etc.
The successful TT passing program has to be able to sort and match
the consensus human conceptual reality to pass itself off as human.

If you plan on creating AI programs you have a lot of work ahead.
indeedy, and I'm hopeless at programming, so it's back to the
cocktail sticks and coloured pieces of play-doh for the time being,

I look forward to your feedback,

n.


I think it is a lot easier to talk about how to do strong AI (mind)
than to actually accomplish anything. Maybe CAs will interest you.

Regards,
Stephen
.

Quantcast