From: Greg A (greg.a@facttechnologies.com)
Date: Fri Apr 07 2000 - 07:44:22 MDT
----- Original Message -----
From: Greg A
To: Marvin Minsky
Sent: Thursday, April 06, 2000 1:28 PM
Subject: Re: Please Call Christopher Alexander @ UCB
Dear Professor Minsky,
I will try to outline the pointers first, and you can ask me to explain more about whatever is behind the ones that interest you most. According to my latest theory on cyborithms, this is shortest path to mutual understanding. It will save a lot of time and wasted reading if you can get the same intuition and follow your own path to the point, and I believe you will within 3-5 *completed* feedback loops of analyzing this concept, so please do reply.
1. I looked up paranomes. I believe I understand something about both their (informal) mathematical and (formal) neurological basis, and I agree that they are the source of metaphoric thought. I would like to apologize again for not finishing SoC before contacting you, but I believe this is urgent, and I wish you to have appropriate credit for the work. I believe it may be possible to build a paranome machine within 6 months to a year.
2. I have read about General Problem Solver indirectly. It sounds as though it operates on cyborhythmic principles. I have heard of work on something called Eurisko by Douglas Lenat that sounds like a superior version. Eliezer S. Yudkowsky (sentience@pobox.com) has gone over it in some detail. The two have had an intermittent dialog on AI topics for the past year or so. As I just discovered, Dr. Lenat is the head of Cycorp. (http://www.cyc.com), and I believe he has been pursuing similar goals. He may be able to add to the conversation.
3. It is immediately important that S-C machines ARE weak. They are useless on their own for exhibiting properties of organizational intelligence such as that exhibited by agents, what I believe (thanks to you) are the units of organizational intelligence. They exhibit algorithmic properties. A *network* of them, however, which can include humans thanks to fact-based modelling, exhibits cyborithmic properties (i.e. problem-solving via cybernetic principles instead of algorithmic). I have tentatively named a machine made up of S-C machines an Artiles-Minsky device, and I hope that in this conversation I can earn the right to keep that title.
4. Full awareness of Bertrand Russell's logical types, and their mathematical analog: human semantic categories a la George Lakoff, is necessary to follow this through to the finish. I hesitate to bring this up, as I am sure you are much more familiar with Russell's work than I am. Nonetheless, I want to be thorough.
5. An S-C machine exhibits some very interesting semantic processing properties, as explained at some length in my online diary entry today (http://www.facttechnologies.com/newworld/diary.html). You may wish to read that first. The "vestigial web content" referenced towards the middle there was the one original insight which led me to this breakthrough -- that *facts* in fact-based modelling can be thought of as *stories* in human thought (as expressed by linguistic tokens).
6. Since you invented paranomes, you have undoubtedly spent some time considering the topological properties of a 3-dimensional, biological neural network. I would also like you to note that biological neurons exhibit cybernetic processing properties because they share the same set of sensors (cell wall neurotransmitter ports) and effectors (neurotransmitter molecules).
7. I have recently been told by an ex-data warehousing student and professional acquaintance of mine, Rick Lowrey (rgl@capgroup.com) that it has been mathematically demonstrated that any n-dimensional system can be properly projected into 3 dimensions, but only special cases can be projected onto 2. This will come into play later, but you may wish to seek some verification of this from fellow MIT faculty. It is more or less hearsay that confirms my intuition at this point. Mr. Lowrey can probably point you to specific research papers on the matter: He has taken an interest in the field and indicated that he has never heard of this math being used for business problem-solving before.
8. I believe the type of projection referenced in #7 is related to Dr. E. F. Codd's 1969 concept of projection in his original work outlining SQL. Codd specifically states that facts can be treated as n-dimensional objects of unit-length edge, an intuition of mine that is reflected in our corporate logo, the pentatope. A review of this work can be found at: http://www.intelligententerprise.com/9811/online2.shtml
9. I believe that #7 and #8 point to a possible mathematical treatment of the idea of a semantic field, which is determined by a set of entities exchanging fact-based data sharing the same (or at least partially the same) set of facts. (See non-video white paper at http://www.facttechnologies.com/newworld for initial outline of the concept of fact-based computing. This was a new revelation for me last Thursday. For that matter, see the "video white paper" on the same page for a relatively quick visual intro to cybernetic principles applied to fact-based data modelling.)
10. I believe that, if a mathematical treatment can be created for the concept of the semantic field, these mathematics will apply *directly* to answering the question of how neurons work together in the brain, and how hive-intelligence emerges in social insects. The physical basis is the exchange of information via visual, tactile, or chemical facts through sensor and effector semantic "gateways" (provided by biology and hard-wired by DNA) in the form of positive and negative feedback loops.
11. As the systems described in #10 are topological systems, the implication is that the basis of human thought and problem-solving is topology-related. Work on spatial metaphor in all human languages (provided by the Lakoff camp) are possible evidence of this, particularly in conjunction with paranomes.
You may wish to pass this around ASAP. The more people that see it, the more likely they will fill in the blanks. With all due respect, sir, I *know* that this is there. The fact that you and Professor Alexander have a relationship via his thesis is probably the source of the root idea.
I'm assuming that this unexpected link is the only reason you wrote me back. The unexpected links in math I get from friends is my basis for my intuitive certainty here.
Thank you again for your time. I am glad you are still alive to see this.
Please let me know if there are any more puzzle pieces I can provide. It's up to the world to put them together. It will take me years to learn the math.
Sincerely, and with Utmost Respect,
Greg A
===============================
Greg A
Chief Executive Officer
Fact Technologies LLC
Giving Business A Mind Of Its Own.(TM)
http://www.facttechnologies.com
877-471-4924 (digital secretary)
510-549-1255 (fax)
greg.a@facttechnologies.com
===============================
----- Original Message -----
From: Marvin Minsky
To: Greg A
Sent: Thursday, April 06, 2000 11:40 AM
Subject: Re: Please Call Christopher Alexander @ UCB
>Dear Professor Minsky,
>
>I do not wish to harass you by continuing to flood your mailbox with
>messages. This will be my last attempt to contact you, but I encourage you
>to call or write me back at any time in the future.
Your strange and intense messages simply do not come to any point. Nowhere do you state what you'd like me to do. In any case, I'm not impressed by attempts to base things on predicate logic, because the assumption of consistency prevents the use of commonsense self-referent reasoning. I explained this in section 27.04 of The Society of Mind.
>I have just faxed a copy of the white paper I sent you a few days ago to
>Professor Christopher Alexander of UCB Architecture. I believe our work at
>Fact Technologies is the link between the work of you both.
In what way? Yes, Alexander has good ideas about combining structural and function representations. (I dimly recall being one of his Thesis readers.) You might find even better ideas if you look up 'paranomes' in The Society of Mind. And it seems to me that, from what you say about S-C machines, these are based on somewhat weaker ideas.
>For my next trick, I intend to integrate the work of Lakoff and Codd to
>produce a machine that understands metaphors. Your support in this effort
>would be invaluable.
You could look up 'metaphors' in SoM.
>Wiener's positive and negative feedback comes into the picture, as you can
>see, on p.2 of the book. Alexander's mathematical treatment of the design
>issues towards the back of the book should be directly applicable to your
>agent theory.
Wiener's feedback ideas were based on the theories of Bode and Nyquist. For AI, they were generalized in the "General Problem Solver" scheme of Newell and Simon.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:08 MDT