Customer Reviews

AI4U: Mind-1.1 Programmer's Manual

5 star:		(0)
4 star:		(0)
3 star:		(0)
2 star:		(1)
1 star:		(2)

 

 

There is no polite way to say this. The author is a crackpot. Reading the materials he provides on his website, an astute reader will notice several things. Firstly, this person doesn't know how to design software at all. He also presumes that his reader knows even less than he does. The "AI" he proposes is nothing more than a basic for(;;) loop. (One of the primitive constructs provided by C, C++, Java and their dirivitives...). He is aware of some of the limititations of his design but is unaware of their obvious (to any intermediate-level programmer) work-arounds. Secondly, while he is happy to put lavish names, such as "Sensorium", on empty or nearly empty functions, he seems to be completely oblivious to the real issues a succesful AI mind must address.

He throws about refferances to concepts in the AI and futurist community such as the technological singularity but fails to demonstrate any understanding of what they mean. He claims that his design solves the AI problem when, infact, it hardly does anything at all.

He claims that his system is suitable for use in robotics, yet he has done no orrigional experamentation.

He continues to troll the usenet (sending between 5-7 messages to every AI and transhumanism related newsgroup per month) pushing his book and his lame ideas.. (If his ideas had even a tenth the merit he claims he would be world-famous...)

I am an AI enthuseast myself and hope to, oneday, publish my own work on the subject. (you can find some of my writings on my website). I do not have the audacity to claim that my work is yet worth anything because I have not yet made much progress. In general, you should stay away from all books on AI unless they are based on actual work that has been done in the field. Work, in this case, being either hard research on biological systems or software development efforts that have shown some type of results.

From any decent developer, you would expect something, which would at least compare to Winograd's SHRDLU. For some reason, the book however is failing to show any results of the "software-architecture" being explained in length!
A scientist's moral dictates, that you withdraw your theories, if you have been proven wrong. Arthur T. Murray however doesn't show any inclination to do so. Instead he is still continuously advertising his long ago falsified book in forums and he's claiming to have developed a powerful approach to AI.
I think, I can safely say, that the author is a phoney and he's only trying to sell his book. So take my advice and don't by it (as I did)!

Murray believes that with the spread of activation
through a network of the correct configuration and
sufficient size you have intelligence and thought.
Not everyone would start from that premise but
probably most connectionist do. While AI4U is
sometimes advertised as a "textbook" it is not
that. An AI textbook should discuss at least the
core AI topics:
search
pattern recognition
knowledge representation
learning
logic
rule-based systems
neural networks
etc.
While AI4U touches on some of these topics it
is not an adequate textbook. Rather it is a
defence of one man's approach to building an
artificial intelligence.
The chapters in this book are too brief and
the discussions too superficial. There also need
to be algorithms provided for each routine in the
code of Appendix A. These could be presented in
pseudocode or as flowcharts for instance.
The biggest problem is the lack of references.
It is just possible that one could write a short
note without finding it necessary to reference the
work of others but it is impossible to write a book
length scholarly work without citing other work in
the field. This is a fatal flaw. Murray should
begin by referencing:
The Structure of Long-term Memory, W. Klimesh,
Lawrence Erlbaum, 1994
Netl, S. E. Fahlman, MIT Press, 1979
Adaptive Information Retrieval, R. K. Belew, U. of
Michigan, PhD thesis, 1986
The authors of these works have accomplished some
of the things Murray is trying to do with Mentifex.
A positive side to Murray's work is that he does
provide downloadable code. When you run this code you
find that Mentifex is very slow even with a very small
semantic network. If one were to build up the millions
of nodes needed to approach human level intelligence
the code would grind to a halt. Murray seems to think
running Mentifex on parallel processors will solve
this problem. I calculate that it will not. I
believe human level preformance requires that one
apply multiple approaches to controling complexity:
category formation by clustering/vector quantization
hierarchical knowledge organization/processing
parallel processing
avoiding search whenever possible
simultaneous use of multiple specialized agents
sequential running of multiple generations of agents
plus any other means you can bring to bear.
(see Asa H, R. Jones, Transactions of the Kansas
Academy of Science, vol 109, No. 3/4, pg 159, 2006)