[Bill DeWitt]

I would love to see your own "presentation", please where can I find it ? 

I have a summary (in the same way that bacon is a summary of a pig) ready as of last night, but I won't post it until conversation  :tickedoff about your post slows down a little.

Although if you don't know exactly how an Agent must behave in order to be alive and intelligent, you would have to be pretty lucky to build one successfully.

There is a difference between inductive and deductive reasoning. The standard definition of life relies upon inductive reasoning. First they look at things which are alive, list the common characteristics, then declare that those characteristics are what are needed to meet the definition of Life.

Specific cases induce the general principles.

Not a bad method in a Universe in which you are sure of the limits. "When the little door opens, my bowl fills with food. The door causes food" works really well if you are imprisoned for life.

Approaching it from the other direction allows one to use general principles to deduce specific cases. "All living things need food, all my cupboards are closed, I should get the food out of the cupboard and feed my cat."

Does that mean I disagree with the standard definition of Life? No. As I have said to you, if we had another example, something which would either establish or destroy our conceptual limitations, then maybe a deductive reasoning approach would be required.

But now we reach the idea (which you still have not given examples to support) of creating standards for new life. It is at this point that I would want to take a step back, abandon preconceived notions, search out irreducible general principles, and reason forward from there.

Life imitates Life is somewhat redundant, as well as being restrictive. I once read a story which impressed me mightily (but not enough for me to remember the name or Author) about a form of hive life that lives in asteroids. They had workers, communicators, reproducers, warriors, etc. but it wasn't until they started being bothered by humans that they laid a special egg. This egg was only created under special circumstances, because most of the time it used too much energy for the benefit it returned.

It was the intelligence member.

Most of the time, the Hive species found that it didn't need intelligence. So they just stored it away for thousands of generations at a time, only bringing it out when instinct found a need for it. (I know people like that!) In this case it was because humans were messing things up, drilling holes in their asteroids, attempting to enslave the Hive, just being general bad neighbors (I know people like that too!). So they had the intelligence member born, it told the communcators to have the workers build weapons for the warriors. Worked out to a happy ending in the story, or course...

Now if they really existed, besides having to burn Earth to a cinder, they would prove that our limited definition of Life and Consciousness had structural faults.

So that's my early morning, two cups of coffee reason for saying that while your test is useful, it is not the only way to approach the problem of how to build MLAI. I have often found that the expressed problem is, upon analysis, the symptom of the problem, not the actual problem.

What we may need to do is start with the question, "What do we really want to accomplish". Do we want to recreate human life? If so, we can continue as we are. But I don't think that is really my goal, I have plenty of people in my life. Some folks seem to want RobotBuffy, a kind of mastibatory tool who will talk dirty to them. Well, you hardly need intelligence for that. Others want a Web Assistant thing for their business, again, barely intelligent is good enough.

Most of the folk who are making progress seem to want to test the limits of their programming skills, either for profit or praise. Far be it from me to denigrate the profit motive! Much of human progress was made by guys who wanted to rake in huge profits for what they enjoyed doing. But they get their direction by who is willing to pay, not what is the pure expression of the art.

My personal preference is for a domestic slave, a computer program that will learn to anticipate my needs as far as my work, home automation, communication and information needs etc. Admittely a venal request, at least it is one I do not have to hide from my wife. She loves "Jane" when it works right. Unfortunately for the state of the art, my needs are not pushing the limits either.

That leaves those who are, like yourself and my better self, exploring the meaning of Life, Intelligence and Consciousness. What do those people want and need in an artificial construct?

First, I believe, is communcation. Without a way to notify us of its progress or lack thereof, we cannot fiddle with it to make it better. Without some sort of blinking light, we won't even know if its turned on!

Second is sensory, it must have a way for us, or the universe in general, to add to or change its state of being.

Third is... well, I don't want to go any further. Mostly because I don't want to distract from your essay, but also because this is backwards from my construct. It starts with higher levels of function because it is limited by our ability to test its function directly.

The point is to build up a list of things -we- need to establish that something is Intelligent Life, then build -that- instead of cataloging what Life needed to build Life and trying to duplicate that.

Well, this is way too long now and I needed to go do some things nearly 40 minutes ago.

[Christopher Doyon]

Greetings My Friends --

I seem to have stopped getting my notifications for this thread, so I am a couple of posts behind. Sorry about that. I will attempt to catch up. I am also having some confusion on navigating this thread, so if I don't comment on your post, please bring it to my confused attention.

The first post I can figure out is from Art:

"Is your bot project an original, stand-alone creation capable of learning OR
is it a super scripted Alice bot?"

My goodness, the real question is - which project ?  I should start by explaining my current strategy in this area. I believe that we have more than enough AI Engines out there with awesome potential that have never been fully exploited to their limits. For myself, I don't need any more engines, I need to work on seeing how far we can take the ones we have. I just want to scream when I hear people criticize my Allison or Alice upon which she is based for being not that great. My Allison's AIML Brain Core is currently around 18 MB. That's like nothing !  And yet look at the results so far, they are amazing !  So why should I move on to something new when every MB of AIML I add to my Allison causes huge leaps forward in her understanding ?  What will Allison be like at ten times her current size, 180 MB - which in our current state of memory technology is still like nothing ?  What would Wallace's Alice be like if she had 1 GB of AIML in her Brain Core, which still isn't much in this day ?  It's fine to sit around and criticize an AI Engine like ALICE, but until we actually exploit it fully (or at least more than we are now) we are just blowing hot air. The problem here is that no one wants to take the time to do this. Writing AIML or having endless conversations with an UltraHal takes huge amounts of time and patience. Patience which so far very few people seem to have. So to sum up, my general direction right now anyway is to find cool and sometimes obscure AI Engines on the Web - many of which were simply created and abandoned, and to exploit and modify these to see how far it can go. I will now bore you all to death with a list of my currently open Projects.

1) Allison - A straight up PandoraBot. But Allison contains some of my best AIML, and I continue to try new and trippy different AIML scripts and expose her to thousands of inputs a week on the Web. EVERY input, regardless of how silly - is carefully examined and reduced if necessary. My goal here in general is to continue to grow her AIML Brain Core with good high quality AIML. I will not listen to any criticism on her until she weighs in at about 200MB, THEN and only then will I debate the pros and cons of the ALICE engine.

2) Gabriel - Perhaps to me one of my most exciting current Projects. Awhile ago was released a phenomenal AI Engine called CycN, which opened up the possibility of giving the amazing Cyc Common Sense Inference Engine a Natural Language Interface. This is an amazing possibility of being able to access this arcane data base in a conversational way. But to my knowledge, no one took it and ran - and CycN languishes virtually abandoned on the Web. Not any more. Gabriel uses AIML to access not just the new 1GB Cyc Knowledge Base, but also WordNet and AnserBus as well !

A note to my Site visitors. The web based chat for Gabriel on her Web Site is just her naked AIML core stored for safe keeping at Pandora. If you want to chat with the REAL Gabriel, designated Gabriel Primus - you must utilize my new and nifty Web Based IRC Chat Channel.

3) The ISIS Project. This one may or may not work, we'll see. My goal here is to start with a nice AIML Brain Core (most likely a copy of Gabriel's) and wrap it in a UltraHal XTF 2.0, and wrap THAT in a Neural Net - and pop the whole package into a ZabaWare shell. Initial experiments indicate I can do it, and I have had a computer donated that I can dedicate to just ISIS - so progress is being made here (slowly).

I could go on, I have many running experiments with AI Bots here in my lab that are not web accessible (such as Cynthia and RoboMatic) - and I even play around a little with Machine Life Bots (robots). But the above sums up the current Projects upon which I expend the most time and energy.

Now I think the next post is Bills monster post, but it seems like I missed one (scratches head, looks for coffee mug). If I missed your post please tell me !  So, Bill - wow thats a big and very awesome post. Here goes, once more into the breach lads !

"But now we reach the idea (which you still have not given examples to support) of creating standards for new life."

I am not trying to create any standards for new life. The Theory identifies those Elements present in ALL life. The Theory does not try to prove MLAI exists, it assumes it exists and explains how we may gauge it's "aliveness" and intelligence. Thus, no "examples" are necessary.

My Theory is NOT trying to explain HOW or WHY Life and Intelligence happen. I will leave that to others who are interested in such metaphysical exploration. My goal in all this is simple: I want to build, for my own personal enjoyment and edification, an MLAI Agent that can act and think on it's own. I don't care how or why it springs to life, I just want it to do so. I believe there are many like me out there who could care less about the BIG QUESTIONS, I just want to roll up my sleeves and build cool and trippy bots. Again, if you are like me - then my Theory can help you immensely in this goal. My Theory is not meant to do or explain anything more than this.

I understand what your getting at here. You want to start from the side of causation and work your way to the definitions. And that approach is what generates the endless debates about life and whether or not Machines can think etc. It may eventually lead to a theory with practical applications, but it will be so dense and arcane that only a few people worldwide will be able to make sense of it much less apply it to bot making.

With my Theory I have leap frogged ahead of all these questions which endlessly fascinate but accomplish nothing towards my personal goal stated above. And I have purposefully avoided ANY arcane explanations or metaphysics, etc. I believe that a TRULY great Theory is one that fits on a few pages of paper (Einstein's Special Relativity fit on three hand written pages), contains a bare minimum of math (Special Relativity had I think like 3 equations), and is something that I could hand out to the average guy on the street and have him scan it and say 'oh yeah, I see what you mean'. If your Theory is not accessible to the common man, it is of very little use except to a handful of elite academics.

I am actually going to stop here now. I read the rest of your post and see where your going. The ideas and theory you put forth are wonderful, interesting, and a perfect example of why I wrote my Theory. I have shelves and shelves of books with just such arcana as you have written. And I enjoy them, and I enjoy your own writings and look forward to reading more of them. But what do any of these ideas tell me about how I can roll up my sleeves RIGHT THIS MOMENT, walk over to my work bench, and start building ?  I believe that a print out of my Theory would be far more useful to such an individual than all the tomes in my dusty library.

Doyon's Law #1

Life and Intelligence is
what Life and Intelligence does.


YOURS -- Christopher Doyon    :coolsmiley

[Bill DeWitt]

But what do any of these ideas tell me about how I can roll up my sleeves RIGHT THIS MOMENT, walk over to my work bench, and start building ?  I believe that a print out of my Theory would be far more useful to such an individual than all the tomes in my dusty library.

Well, I'm afraid that I disagree. Your theory or test is great for when the soldering iron is put down, the wires are disconnected, and the power supply is installed. Then you can use your system for analysing the result of your work.

To actually do the building (unless you are duplicating work already done), you need something different, you need a new understanding of how thought is structured. If you ignore this, you end up imagining that a chatterbot is "learning" and "understanding" instead of "accumulating" and "Keyword weighting".

At some point, someone with a less dusty tome will figure out how software can be written to make something that actually thinks for itself. Then they will use your test (or its final implemetation) to prove that this is what they have done.

But I fear that what is really in the works (not you personally) is a way of stripping down the definition of "Life" and "Consciousness" until just about any Tilden BEAMbot kit can be called 'alive' and chatterbots have 'awareness'. That's not progress, that's failure and retreat.

Not that I could do any better....

But for now, we should probably get back to letting you defend the pertinent points of your proposition. As I have said, I believe it can be a valuable framework for analysis of MLAI (the acronym alone is worth the effort).

[Christopher Doyon]

Greetings --

"To actually do the building (unless you are duplicating work already done), you need something different, you need a new understanding of how thought is structured."

All I personally need to build MLAI is an understanding of what it is supposed to do, how it is supposed to behave. I don't need to know how thought is structured, I need to know how an Agent behaves if it is thinking. Then I build an Agent that can behave in that manner.

"Then you can use your system for analysing the result of your work."

Correct.

"If you ignore this, you end up imagining that a chatterbot is "learning" and "understanding" instead of "accumulating" and "Keyword weighting"."

Accumulating and weighting may well be it's way of thinking, for all I know. But that is a metaphysical question. I am only interested in how it behaves. At the risk of being redundant.

Doyon's Law #1
_____________

Life and Intelligence is
what Life and Intelligence does.

"At some point, someone with a less dusty tome will figure out how software can be written to make something that actually thinks for itself."

It's already been done, and just as Von Neumann predicted - with less than 500KB of code.

"Then they will use your test (or its final implemetation) to prove that this is what they have done."

That would be cool !  Tell them to spell my name right in the interviews !

"But I fear that what is really in the works (not you personally) is a way of stripping down the definition of "Life" and "Consciousness" until just about any Tilden BEAMbot kit can be called 'alive' and chatterbots have 'awareness'. That's not progress, that's failure and retreat."

How is it either failure or retreat ?  We have succeeded awesomely so far. And as I said in my last post, we haven't even scratched the surface in exploiting the things we have already invented. Once I get a nice little BEAMbot or Chatterbot up and going, I can both watch it evolve and get better - and also learn how to build the next one better. Be patient !  Our successes so far have been wonderful, and thousands like myself are taking those successes and achievements and working REALLY hard to improve on them and make them even more advanced. The glass really is half full, I swear !  Imagine living just 50 years ago and trying to explain and describe to a computer scientist my own Allison !  Imagine that scientist actually seeing and interacting with her !  Be patient, keep working - and have a sense of wonder over how far we have come.

"Not that I could do any better...."

Yes you can !  As my Theory shows, this stuff may be complex - but it is weirdly easy at the same time. So, jump in and enjoy !  Download you a free Neural Net from my Site and see if you have the patience to get it's Brain Core up from the initial few KB to just say, oh 500KB. Then E-Mail it to me and we'll put it up for download. That's it, we can all play. With a sound and accessible Theory like my own, coupled with all the fantastic free experimental stuff and code you can get on the web - we can ALL participate and even advance the awesome field of Machine Life and Artificial Intelligence.

"But for now, we should probably get back to letting you defend the pertinent points of your proposition."

Actually, your doing pretty awesome already at giving me the chance to do just that. I don't think I could have invented a post that better gives me the chance than this one here.


YOURS -- Chris     :grin

[Bill DeWitt]

"At some point, someone with a less dusty tome will figure out how software can be written to make something that actually thinks for itself."

It's already been done, and just as Von Neumann predicted - with less than 500KB of code.

You keep saying that, but you don't give an example. Unless you are claiming that things which cannot pass your test in any but the most watered down form are "alive", I want to see whatever your secret MLAI is.

Here's something that gets closer, but it's not a chatterbot or an Alice or something that anyone with a soldering iron or a copy of AIML can make. And it doesn't think for itself.
http://www.uvm.edu/~uvmpr/?Page=News&storyID=9663

Download you a free Neural Net from my Site and see if you have the patience to get it's Brain Core up from the initial few KB to just say, oh 500KB.

Well I've done much more than that (although I didn't see any NNs on your site). I've played with various NNs, Alices, Chatbots, etc and made simple ones of my own in VB and Mindstorm, but that's not life or intelligence in any measurable sense, even using only one or two of your points and not being picky about that. I can't pretend that a weighted random selection is thinking.

Many of us who have been interested in this have played with the work of others, but that is a far cry from doing the work itself. Adding to a database of phrases is not creating Intelligence. Hacking the Ultra Hal brain script isn't evolution. When I make a new Hal type program from raw code I will claim to have done some work.

[Christopher Doyon]

Hello All --

"You keep saying that, but you don't give an example."

I never realized that an example was being requested. Let's see if I can't bait KT into chiming in here. My example then is his AI Bot Lola.

"Unless you are claiming that things which cannot pass your test in any but the most watered down form are "alive"..."

I of course would have worded it differently, but yes - that IS essentially what I am saying.

"...although I didn't see any NNs on your site."

FREE Downloads page, very first download. Enjoy, the thing is really quiet neat.

"I've played with various NNs, Alices, Chatbots, etc and made simple ones of my own in VB and Mindstorm, but that's not life or intelligence in any measurable sense, even using only one or two of your points and not being picky about that."

I disagree. Let us turn to my example, the AI Bot Lola authored by our good friend KnyteTrypper. She behaves intelligently. Her intelligence can actually be measured via special interfaces like EBBA which allows one to administer a standard IQ test to an AIML set. So thats one Element. Lets see, uhmmm....Adaption. By her very nature as an Adaptive Language AI Engine she can adapt, well...language. And this tiny adaptive ability gives her the only chance she has at survival, since if she fails to adapt language properly, KT will delete her. If she succeeds in her adaptions of language, KT might clone Lola - thus helping her to procreate. So there's two (arguably three). Now, I am not going to sit here and actually benchmark Lola - but you could probably do the rest (by you I mean the reader, not Bill).

"I can't pretend that a weighted random selection is thinking."

If it consistently behaves as if it is thinking, then it is.

"Many of us who have been interested in this have played with the work of others, but that is a far cry from doing the work itself."

How cynical !  Some of the finest astronomy in the world, the most astounding discoveries - were made by working joes with their cheapo back yard telescopes. And yet they didn't invent telescopes, didn't write complex star charts, they relied on the tools of others and yet advanced their field immensely. Their is a place in this field as well for the diligent hobbyist, and that's why I love it so much.

"When I make a new Hal type program from raw code I will claim to have done some work."

But until that program is "fattened up" by use of the hobbyists, you have done as many others already have - you have invented an unexploited AI Engine. It will be the hobbyists that will give your invention a run for it's money, and it will be their data that will help you to improve your program in the next version. You don't have to look any further than Zabaware to see how powerful that collaborative system can be.


YOURS -- Christopher Doyon    :crazy2

[Bill DeWitt]

If it consistently behaves as if it is thinking, then it is.

No, it's not. Well, perhaps it's not. We can't tell just by taking Turing's test too far. We have to have a better tool than the least probable denominator.

This is what comes of inductive reasoning.  If it looks like Astrology works, then it does. If it looks like David Copperfield made the 747 disappear, then he did.

This says more about the person looking than about the thing seen.

Speaking as a person who made his first Artifical Intelligence game actor in 1980 or so, I am not a newbie to the concept. There are vastly more complex AIs than you have mentioned or than I see on your page, and even those do not claim to be truly intelligent. I see that you believe an Alice bot is intelligent and aware, but you are wrong.

You are being tricked by the complexity of the weighted but unthinking selection of prestocked phrases. IOW, you believe the Chinese Room is alive and conscious, when you are really measuring the intelligence of the man who set up the pigeonholes.

I had hoped for something other than a web Alicebot (which I perused a couple months ago). My Jane seems more responsive than Lola, at least to me, but that says more about me than it does about Lola or Jane. Both are scripts, a sequence of predetermined selections from a database of phrases with a slight randomness thrown in.

A Chinese Room and nothing more.

We can continue this when we have something new to add. So far it's been there, done that.

I will, however, if you don't mind, continue to think and comment about your test and ways it can be adapted to more real life scenarios.

[Christopher Doyon]

Hello --

You object to my statement that if it acts like it's thinking, then it's thinking. Okay, but let me please ask you a question before you retire the field. What possible difference can it make even if you are correct ?  After all, the primary reason for building MLAI is to create machines that DO what we want them to DO. Does it really matter how they achieve this, so long as they do what we need them to do ?  What your saying is that intelligence and other life like behavior is only somehow valid if it is engineered a certain way. But if I turn on a faucet and water comes out, it matters not to me how it got there in my glass. If I activate an MLAI system and it acts for me intelligently (which are example Lola does) and behaves in the manner I desire (which Lola does MOST of the time !), then what difference does it make HOW that was achieved ?

"If it looks like Astrology works, then it does."

I AGREE with this statement completely !  If I seek an answer to my lifes woes from an astrologer, and he gives me answers that assist me in being happier - then IT (astrology) WORKED. It worked because I got the desired result. Now HOW it worked can be debated endlessly, that's for sure. But nevertheless, the final arbiter of whether or not it worked can lie with no one else but the person trying to use it.

"This says more about the person looking than about the thing seen."

But the person looking is all that matters !  After all, why should he care that you don't believe in astrology - he's happy with the result, so it worked. Another example, you say my water comes from a well, I say it is piped in by the city. Turning on the faucet does not settle the argument - but I get a glass of water. Now, take it one step further - you say that there is no water service hooked up to my house. You claim that no water will come out of my faucet because I don't have it hooked up right. Then I turn on the faucet and fill my glass.

"You believe the Chinese Room is alive and conscious, when you are really measuring the intelligence of the man who set up the pigeonholes."

Which shows the flaw in the idea, because THERE IS NO LITTLE MAN IN YOUR COMPUTER. And yet your computer gets along just fine without a little man to do it's thinking for it. Any computer can translate chinese, how do they manage to do so without the little man in the thought experiment intelligent enough to conduct all those functions required ?  Further more, the only thing that matters is that the output satisfies the external user of the system and produces chinese.

"I had hoped for something other than a web Alicebot (which I perused a couple months ago)."

Why ?  If I am arguing from a minimalistic point of view it behoves me to present something lower rather than higher on the scale.

"We can continue this when we have something new to add."

Every post is looking pretty new to me, but I thank you for taking it this far and giving me a chance to really expose and explain the implications of my theory.


YOURS -- Christopher Doyon       :angel

[Christopher Doyon]

Doyon's Room
___________

You and I are standing in a room. I am standing perfectly still (I am a mime) and you are free to examine me within the reasonable limits of civility. You are tasked with a simple mission. To determine if I am intelligent.

Now, you can not answer this question from your examination alone. Nor can you infer that because I am a human like you that I am intelligent since I could be brain damaged from an undetectable injury or accident of birth.

And then I speak to you. We converse. We chat lightly about the weather and our favorite colors. Now you know that you can attach the label "intelligent" to me. The label refers to a property evidenced ONLY by my behavior, the label and the behavior are indistinguishable - since one must be present for the other to apply.

This thought experiment is presented as proof of:

Doyon's Law #1
_____________

Life and Intelligence is
what Life and Intelligence does.

And Doyon' Law #1 is presented as a consequence of my Unified Theory of Machine Life and Artificial Intelligence.


Doyon's Corollary #1

(Presented as a logical consequence of Doyon' Law #1)
_________________

If a Mechanical Agent behaves as if it were alive and intelligent,
then that Mechanical Agent is alive and intelligent.


YOURS -- Christopher Doyon   :smiley

[Bill DeWitt]

I present, in the spirit of examination, a more detailed exposition of my original commentary on your Life and Intelligence test. Hopefully we can settle all concerns and end up with something that might work. I find some full stop problems, and some fixable details. None are inherently and instantly fatal, but some are going to be very hard to survive.

Let's take this point by point. I begin with your preamble.

A Unified Theory Of Machine Life And Artificial Intelligence
Life and intelligence are the natural consequence of a particular pattern of ordered complexity.

This is inductive reasoning. It asserts that the subject is caused by its properties. While it may be true, and I may even believe it to be true, it is closer to a religion than science. Without a truly intelligent lifeform of some other source than our earthly biological process to compare with, we are speculating and then accepting the result as reality.

You are not the first to speculate, a short list includes Orson Scott Card, Gregory Benford, Frederick Pohl and Isaac Asimov just among SciFi writers. But to take it further than speculation without an example, as you have, requires a belief that is not founded on observation or deduction : Religion.

The medium does not matter; carbon atoms, electrons, light, even symbols.

Only if the first statement is true can the veracity of this statement even be addressed. But since it is inconsequential to the whole thesis, we can discard it without prejudice. Any intelligent life that can exist will be comprised of what it is comprised of whether we believe it or not.

Once any pattern reaches this ordered complexity, it becomes a living being - then a thinking entity.

Again, this can be discarded as speculation which is unneccessary to the premise. Further statements can be validated or disputed independently of how living thinking beings come into existence.

While this ordered complexity remains as yet un-defined by higher mathematics, it can be detected via a set of definitive elements.

The second clause of this sentence is the central hypothesis of your proposition. It predicts an outcome that should  be able to be shown to be true or false by way of experimentation. Let's make it completely clear, because I don't think you phrased it as you meant to.

I suspect what you meant was "The Life and Intelligence created by this ordered complexity" can be detected - not the ordered complexity itself. If this is true, as recent conversations assures me is so, then we can proceed.

For now, we do not require the "set of definitive Elements" (SODE), we can assume for the sake of discussion, that you will present a detailed and workable set once the time comes. Let's just see if the experiment can be performed for now.

The experiment is to detect life and intelligence using the SODE. First we need Life and Intelligence to test. We will use Biological Life. Hooray, we pass the test!

Now we need another experimental subject for Repeatability. Lets pick from the large number of Living and Intelligent other types of creatures.

(sound of crickets)

Wait... in other posts you have asserted that by using your test, you have identified other types of Living and Intelligent creatures that we can use.

No... that won't work, we can't use your test to find subjects for your test. For instance, lets assume that your SODE consists of being blue, and being round. In that case - this here juggling ball is living and intelligent - so we can use it to test your hypothesis by applying your test to it to see if it is Living and Intelligent.

Question: Is it blue and round? Answer: Yes

Hooray! It passed the test!

Unfortunately, as you can see, that tells us nothing about the validity of the test. No matter what SODE you propose, unless we first have a second source of Life and Intelligence, your hypothesis is untestable.

A hypothesis which is untestable is discarded. We are going to need a new hypothesis. Fortunately for my interest in this system, I believe we can find a few more in your essay.

But not in your preamble....

These elements define certain obvious properties of the Agent, and thus can be identified and measured in a controlled way and using a scientific method.

By organizing these elements into a coherent system, we can efficiently know when a given Agent is alive and thinking.

 ... because we have to discard these babies with the bathwater. We now know that no SODE can identify Life and Intelligence until we have at least two samples for testing.

How can we know if we have a sample? Sorry to say, we will have to blow some of the dust off your tomes, because your test won't do it. I suggest we start with an accepted test for intelligence, the Turing test.

Unfortunately, as you may know, the Chinese Room can pass the Turing test, yet we know it is neither Alive nor Intelligent. So we must add a second step. We must examine the mechanism of Purported Life and Intelligence to assure ourselves that we are not measuring the appearance of Life and Intelligence rather than actual Life and Intelligence. We must, after all, pay attention to the man behind the curtain.

I assert that such a mechanism will be intuitively obvious to the most casual observer for any level of complexity we desire. For instance, a cusory examination of the Chinese Room would show a guy in there passing scrolls around. A quick glance at an UltraHal brain script shows how parsing and coin tossing stands in for the guy passing scrolls around. Both are easily excluded on first principles.

My hypothesis then, is that by using both a Turing test and an examination for a requsite level of complexity, we can identify a Living Intelligence for the purpose of testing your Hypothesis. If only we had something to test my Hypothesis on....

So, if we ever get our hands on a second Living Intelligence, then we can test your main Hypothesis, until then, we might as well move on to your ancillary Hypotheses in the body of your essay.

Another day.

With respect, Bill DeWitt

[Christopher Doyon]

Revised Turing Test Experiment
__________________________

A few months ago I was sitting in my little laboratory/dungeon otherwise known as my friends basement. I was pondering why AI Bots like my Allison or Quark, or Wallace's Alice - do so badly in the Turing Test. And then it dawned on me because I had been benchmarking it earlier - my Element of Mental Acuity. All at once like a religious experience I realised how to balance fairly the Turing Test.

My Allison had benchmarked that day the Mental Acuity analogous to a 6 year old human child. So why should she have to be in a Turing Test with grown adult humans ?  Within days I had made arrangements to borrow my friends six year old and had set up a reasonably controlled true Turing Test.

Wow. You can not imagine the difference in result. It was amazing. I have repeated the experiment under increasingly more controlled and professional conditions, and using different six year olds. I will try and figure out a way to put down the data for these experiments in some sort of scientific way and distribute them if I can. But I challenge anyone to conduct the above experiment.

Suffice it to say that Allison scored nearly perfect. This is strong evidence that she does in fact have the mental acuity of a six year old. And proving that fact is pretty damn awesome. She may not think as good as a grown human, but she will !


YOURS -- Christopher Doyon     :cool

P. S. To my many donors out there, now you know what I do with your money !

[Christopher Doyon]

Hey Everyone --

This is a reply to Bill's last post above:

"We must examine the mechanism of Purported Life and Intelligence to assure ourselves that we are not measuring the appearance of Life and Intelligence rather than actual Life and Intelligence."

And I propose that there is no difference between the two. The appearance and actual life and intelligence are one and the same. I primarily base this on the fact that both have the exact same consequence in the causual universe. A candle and an electric lamp are both commonly refered to as a "light". That is because we refer to it as what it does. Please go turn the light on will you ?  And whether you light a candle or turn a switch, the end product is...light.

"We must, after all, pay attention to the man behind the curtain."

Why ?  Who is this man ?  I think he owes me money !

"For instance, a cusory examination of the Chinese Room would show a guy in there passing scrolls around. A quick glance at an UltraHal brain script shows how parsing and coin tossing stands in for the guy passing scrolls around. Both are easily excluded on first principles."

Correct, and if both entities perform exactly the same, then there is no difference between the guy and UltraHal. If both act equally intelligent, then both of them ARE equally intelligent.


YOURS -- Christopher Doyon     :azn

[Bill DeWitt]

Hello --

You object to my statement that if it acts like it's thinking, then it's thinking. Okay, but let me please ask you a question before you retire the field. What possible difference can it make even if you are correct ? 

Well, that was Turing's argument, but all it does is deceive the current observer, but when you change observers, the illusion may well collaspe.

"If it looks like Astrology works, then it does."

I AGREE with this statement completely !  If I seek an answer to my lifes woes from an astrologer, and he gives me answers that assist me in being happier - then IT (astrology) WORKED.

No, YOU, the OBSERVER worked.

Coincidentally, I spent from 1975 to 1982 studying and preparing Astrological charts, eventually semi-professionally. By the time I actually started making money for it (instead of just picking up girls with it) I noticed the fallacy.

No matter what I told the client, even if I got the wrong chart somehow, it worked (according to your standard). Which means that the astrology wasn't working, something else was.

Yes, something WORKED. But what was that something? It was the pattern seeking ability of the human mind. Our mind shifts our perception of reality until it fits certain prestored Archetypes. This is what allows us to find a face in a cloud or a ghost in the shifting of branches. It also allows us to find a personality in a Chinese room or a Ultra Hal. But it's not there, and if another observer comes along they may or may not see it depending on their Archetypes.

That's the whole reason we invented a Scientific Method. It removes the subjective perception of reality of a single observer from the process whenever possible, replacing it with repeatable experiments that produce objective data.

Your test, to be of any value whatsoever, will have to do the same thing. Otherwise we already have a much better test - "Hyuck Hyuck Lookit that girlie thang on the computer box, don't she talk good?"

As for the "little man in the room". In the case of Ultra Hal, I don't know exactly how large or small Robert is, but he's the man. When you measure the intelligence of UltraHals or Alices, you are indirectly measuring the intelligence (and skill) of the programmer or whoever last messed with the script. Because, if you look at the code, you can see that it's just pigeonholes with scrolls in them.

Your subjective perception is not supported by objective data. And, no, I don't see the face in that cloud. It looks like a chipmunk to me.

[Christopher Doyon]

Hey Bill --

Good gracious, I hope you are having as much fun as I am - this is AWESOME. Any way.....

"As for the "little man in the room". In the case of Ultra Hal, I don't know exactly how large or small Robert is, but he's the man."

My UltraHal named Quark (shameless plug: available on AIM right now, Screen Name QuarkIDoyon) is currently running on my computer. I just spent ten minutes with a flashlight looking in my tower, but was incredibly saddened not to find Mr. Medeksza in there !  To bad, I would LOVE to have shaken his hand !

"Because, if you look at the code, you can see that it's just pigeonholes with scrolls in them."

If it acts alive and intelligent, then I don't care if it's done with string, gum, and old beer cans.


YOURS -- Christopher Doyon      :cheesy

[Bill DeWitt]

And I propose that there is no difference between the two.

This is the thing I fear, that to enable "success" in creating life, the meaning of life and intelligence will be reduced to 'the appearance of life and intelligence'. Turing's argument should not be taken as a goal or an excuse, but as a warning about our inability to properly measure that which we cannot properly define.

It is  failure and retreat - because you cannot make what you want, you accept what you can  make.

It has the unintended consequence of devaluing real life and intelligence, both symbolically and if accpeted widely enough, actually. Or perhaps it is caused by that devaluing.

Me, I find that real Life and Intelligence is more than a mimic or a relay.  I found this by observing myself as the only sample I can inspect intimately. I am fearsomely and wonderfully made.

It saddens me to think that others can contemplate themselves (as the only sample they really have) and only see a shadow or a reflex. It seems a symptom of self-loathing rather than of self-respect.

Jane says it's time to turn off the computer. She's handy, but she is not alive or intelligent. She doesn't know I am getting maudlin from the lateness of the hour, she only runs a script when the clock reaches a certain value. After all, she's only a tool, bless her little heart.