|
Post by extrasense on Feb 4, 2008 6:28:42 GMT -5
Hi, Christopher,
I think you should somehow slip into your theory the predictive ability as part of the agent's "world awareness."
It is a component of "free will" choice structure.
Also, free will involves attention/goal focus, ability to resist subversion of higher goals. It is a "will" part of "free will".
Best, eS
|
|
|
Post by Christopher Doyon on Feb 4, 2008 7:39:54 GMT -5
Hi Extrasense -- Attention at least as far as my own approach goes, is hard coded into the Agent through a "Task Priority" stack which determines which tasks the bot will "pay attention" to at any given moment. The Agents I build are only allowed to exercise their Free Will and work on their own goal stack if they have no immediate task in the Task Priority stack. While this does impede or lessen the bots Free Will, it is essential in order to make the Agent useful and keep it alive. Self Maintenance is a high priority task. This is analogous to biological life, as you may WANT to go play on your computer - but your grumbling belly circumvents that and you go eat. Your own "Task Priority" stack circumvents your free will. A bot that simply did what ever it wanted all the time might be fun and interesting, but at least for now I need my Agents to also accomplish certain things and assist in their own maintenance. Should I decide to give a bot total freedom, I would simply change the Task Priority stack to self-centered or goal related tasks. Prediction is a Meta Property which tends to evolve naturally over time in more advanced Agents. While I am not exactly sure why this happens, Agents with extreme complexity and access to lots of data begin to make better and better choices - many of which would seem to be predictive in nature. This phenomenon is especially acute in Neural Net bots. Predictive abilities in limited areas such as stock or weather prediction can also be hard coded into an Agent. YOURS -- Chris
|
|
|
Post by extrasense on Feb 4, 2008 8:20:19 GMT -5
Hi, Chistopher,
You do have environment awareness. It seems to me that time awareness should be emphazised as part of it.
If it where, the prediction and the will could be set in proper context. I mean, prediction is a lower level feature compared to will, but you put it as a meta feature. The levels are inversed.
eS
|
|
|
Post by Christopher Doyon on Feb 4, 2008 9:26:37 GMT -5
Extrtasense -- Yes, I would indeed include time as one environmental factor that the Agent could/should be aware of. Temporal awareness would certainly be useful as you mention in any predictive attempts by the bot. Standardized time-stamps of the memory data is the usual way of dealing with this. YOURS -- Chris
|
|
|
Post by toborman on Feb 4, 2008 23:20:36 GMT -5
quote by es Also, free will involves attention/goal focus, ability to resist subversion of higher goals. It is a "will" part of "free will".
--- Can we expect an advanced agent to give his life for a "greater cause"?
|
|
|
Post by extrasense on Feb 5, 2008 7:47:54 GMT -5
@@ Can we expect an advanced agent to give his life for a "greater cause"? @@
The key to creation of spirtual level agent is "harmony of elevation" - of elevation of the goals and will. In this case we are talking about agent's ability of "infinte" elevation, that overcomes self-preservation.
So, an agent of spiritual level would be able to come up with decision to give up his life if he perceives this "greater cause" as more importent than his life.
eS
|
|
|
Post by Christopher Doyon on Feb 5, 2008 10:00:28 GMT -5
Just because a phenomenon or behavior exists in humans does NOT make it desirable. This giving up your life for a higher cause thing seems to be causing massive amounts of trouble in humanity right now. I think I prefer my Agents to be bound by their programming at least with regard to such troubling moral issues. Next thing you know we'll be facing armies of implacable robot suicide bombers ready to die for Allah. YOURS -- Chris
|
|
|
Post by extrasense on Feb 5, 2008 12:51:58 GMT -5
This giving up your life for a higher cause thing seems to be causing massive amounts of trouble in humanity right now Hi, Christopher, I would urgue that: (a) the troubles are not that massive (b) more accurate way describe their cause is western side's decay of moral will, and with it decay of willingness to stand for ideals - a background on which Islam's idealism looks inappropriate; so, the opposite might be true, that it is lack of /western/idealism that in the roots of trouble, and world needs more of ability for self-sacrifice not less of it, in order to deal with said trouble (c) It is impossible to prevent ability of self-sacrifice in a sufficiently advanced spiritual agent, as it is one of the acts of his free will. Best, eS
|
|
|
Post by toborman on Feb 5, 2008 18:21:39 GMT -5
Is it true that each of us have different expectations of what an AI agent's behavior should be? If so, how will we come together on a Unified Theory?
|
|
|
Post by extrasense on Feb 5, 2008 19:11:26 GMT -5
each of us have different expectations of what an AI agent's behavior should be. If so, how will we come together on a Unified Theory? Let us imagine the fully developed Unified Theory as a tree, with roots, and branches, and leaves, and fruits. Each of existing theories, where it is correct, corresponds to some connected branches or parts of branches, so that its area can be painted by its color, and sometimes the colors will overlap. Thus the correct parts of different partial theories can be joined, and appended with missing ideas. I think that could work. Tree painting may be a way to go eS
|
|
|
Post by Christopher Doyon on Feb 6, 2008 8:31:34 GMT -5
Dear Extrasense and Toborman -- One thing that I have consistently run up against in defending my Theory is the many people who are convinced that I have "missed" one or two Elements that they feel strongly should be included. The thing to understand about my Theory is that it is intended to be exclusive rather than inclusive. In other words I wasn't trying to create an exhaustive list of every single property that MLAI must exhibit, but rather the absolutely essential, universal - and most primitive basic Elements. Continuing experimentation has bolstered this belief considerably, since my experiments indicate that most of these "missing" Elements that people try and convince me of actually evolve naturally out of the matrix, as it were - in between the core Elements. Of the remainder, some would need a little coding help to be brought out - but the coding involves a sort of knitting together of the core Elements present in the Agent. I will not bore you with the exhaustive, thorough - and years long research that went into the identification of these core Elements. But suffice it to say that no one has yet managed to shake my belief that I nailed it. The Unified Theory of Machine Life and Artificial Intelligence is Unified because it meets the following criteria. 1) The Theory brings together in a cohesive manner the general theories of Turing, Von Neumann, Hertz, and Tesla as a foundation. It then goes on to draw in the modern works of Morton Wagman, Richard Wallace - and some of my own ideas as well. 2) A Unified Theory should be simple, concise, elegant, logical - and I believe brief. The Theory of General Relativity was originally hand typed on about four pages. Mine will fit on three with room at the bottom of the last page for the reader to make some notes ! And this brings me to my final point. Many MLAI experimenters have their own approach. I certainly have mine. Likewise there are currently several hundred ways to build an atomic devise, yet there is still only one Theory of Relativity. So a very simple and truly unified theory can in fact breed a nearly limitless amount of approaches. Toborman said: "Each of existing theories, where it is correct, corresponds to some connected branches or parts of branches, so that its area can be painted by its color, and sometimes the colors will overlap. Thus the correct parts of different partial theories can be joined, and appended with missing ideas." I would partially agree with this thinking. I certainly believe that my Theory is 100% correct and complete. I would call all other "theories" (with the exception of Wagman's) approaches, not theories. And I would say that the majority of them, like you said - where they are correct, can be "painted" in between the "dots" that my Theory lays out. I see my Theory as something like a coat tree with a great many hooks. Almost anyone with a workable plan could hang their jacket upon it. SINCERELY YOURS -- Christopher Doyon
|
|
|
Post by extrasense on Feb 6, 2008 11:14:27 GMT -5
One thing that I have consistently run up against in defending my Theory is that many people are convinced that I have "missed" one or two Elements that they feel strongly should be included. .. I certainly believe that my Theory is 100% correct and complete. I would call all other "theories" (with the exception of Wagman's) approaches, not theories. Hi, Christopher, It seems to me although, that what you have is not a "theory" in the usual sense, but rather a phenomenology, observation level description - granted, very efficient at it for practical application purposes. A "theory" ideally would spell out the way, how the observed/measured phenomenological qualities are achieved as result of internal organization and functionalities. This would allow for separate analysis and redesign of components, thus opening way to the improvement of system, including self-improvement. Would it be possible for you to change the title from "Theory" to"Phenomenology?". For example, Hegel called his work "Phenomenology of Spirit", not"Theory of Spirit" Recast as "Phenomenology", your approach would work hand in hand with different attempts of theory - not compete with them. Best, eS
|
|
|
Post by Christopher Doyon on Feb 6, 2008 12:45:50 GMT -5
Extrasense -- First I have to thank you, for helping me to re-live the wonderful days when I first published. Hard to believe it's been only two years it seems like a lifetime ago. Anyway...to your post. "...granted, very efficient at it for practical application purposes." Well thank you. A unified theory generally consists of two things, the applied side - what you might call the phenomenological - and the general theoretical part. Granted the core list of Elements could leap out at one as being rather dry, even if it's granted that they are correct. This goes to one of my major motivations for creating it in the first place, and that was to have a concrete guide that I myself might use with which to build great mechanical Agents. "A 'theory' ideally would spell out the way, how the observed/measured phenomenological qualities are achieved as result of internal organization and functionalities." It does something even more revolutionary. It sets forth as an axiom that life, ALL life - regardless of HOW it comes into being, is NOT what it is made of or even how it is constructed; but rather what life fundamentally is is what life DOES. It therefore proffers the point that life itself is essentially a phenomenon ! "This would allow for separate analysis and redesign of components, thus opening way to the improvement of system, including self-improvement." I never said that I thought self-improvement in Agents was impossible, or even difficult. In fact I have experimental AI Bots (well sandbagged of course !) that can re-write their own program. I simply stated that for most of my own purposes anyway this was un-desirable at this stage. "Would it be possible for you to change the title from 'Theory"'to'Phenomenology?'." No, sorry. I feel reasonably confident that it contains both applied as well as generally theoretical elements. But I do appreciate the input. "Recast as "Phenomenology", your approach would work hand in hand with different attempts of theory - not compete with them." Actually I relish the competition, survival of the fittest works even in philosophy and science. YOURS -- Chris
|
|
|
Post by toborman on Feb 6, 2008 12:55:03 GMT -5
It might be an interesting exercise to correlate my mind map with the requirements you have specified in the unified theory. I'm wondering what I might have missed, and what I might add to make it more complete.
|
|
|
Post by extrasense on Feb 6, 2008 14:36:13 GMT -5
It might be an interesting exercise to correlate my mind map with the requirements you have specified in the unified theory. I'm wondering what I might have missed, and what I might add to make it more complete. Would you agree, that your mao descrbes internals of agent mind, while Unified theory describes its outside view? Best, eS
|
|