MISTIC: 350,008 Item Pairs
Christopher McKinstry
MISTIC (Minimum Intelligent Signal Test Item Corpus)
http://www.clickable.com/mist_corpus.html has passed 350,000 item pairs
as of March 22, 1997.
The MIST Item Corpus Project is the largest (in terms of participants)
Artificial Intelligence project in human history. This is due entirely
to the WWW and people like you. The goal of this project is to make a
high resolution conceptual map of human cognition. This map will be used
in various research projects and might well become the basis of an
artificially conscious computer system.
Please visit MISTIC and contribute your own items, such as these:
Beds can be used to sit on. :true
When gardens are in bloom they attract more interest. :true
All dimensions of a container must be considered to dermine its volume.
:true
At birth, human beings can walk on two feet. :false
Orange juice tastes bad, for a short time, after brushing your teeth.
:true
Most people sleep with their eyes closed. :true
Swimming with ski pants on is more difficult than swimming nude. :true
In 1977, the Earth revolved around two suns. : false
--
K. Christopher McKinstry
Homepage: Http://www.clickable.com/employees/chris
World's Largest AI Project: Http://www.clickable.com/mist_corpus.html
World's Coolest Interactive Drama http://www.cr6.com
Jim Balter
Christopher McKinstry wrote:
>
> MISTIC (Minimum Intelligent Signal Test Item Corpus)
> http://www.clickable.com/mist_corpus.html has passed 350,000 item pairs
> as of March 22, 1997.
>
> The MIST Item Corpus Project is the largest (in terms of participants)
> Artificial Intelligence project in human history.
And the least intelligent.
> Beds can be used to sit on. :true
Tell that to an elephant.
> All dimensions of a container must be considered to dermine its volume.
> :true
A great way to figure out the volume of a trombone.
> In 1977, the Earth revolved around two suns. : false
With questions like this, an intelligence, artificial or not, might
wonder why they are only up to 350000.
--
<J Q B>
John Nagle
Jim Balter <j...@netcom.com> writes:
>Christopher McKinstry wrote:
>> MISTIC (Minimum Intelligent Signal Test Item Corpus)
>> http://www.clickable.com/mist_corpus.html has passed 350,000 item pairs
>> as of March 22, 1997.
>> The MIST Item Corpus Project is the largest (in terms of participants)
>> Artificial Intelligence project in human history.
>And the least intelligent.
>> Beds can be used to sit on. :true
>> :true
Well, the "Cyc" database is rather like that, although better
organized for machine processing.
John Nagle
Christopher McKinstry
Jim, MISTIC has a very simple purpose... build a test that can tell the
difference between a human and a random system with a very high degree
of accracy. Every single question that has a high between human
stability is important.
This project has been set up to continue to collect data long after both
you and I are dead. It's bigger than me, it's bigger than you, and more
important than either of us.
History will not be kind to your attitude if it remembers you at all.
Jim Balter wrote:
>
> Christopher McKinstry wrote:
> >
> > MISTIC (Minimum Intelligent Signal Test Item Corpus)
> > http://www.clickable.com/mist_corpus.html has passed 350,000 item pairs
> > as of March 22, 1997.
> >
> > The MIST Item Corpus Project is the largest (in terms of participants)
> > Artificial Intelligence project in human history.
>
> And the least intelligent.
>
> > Beds can be used to sit on. :true
>
> Tell that to an elephant.
>
> > All dimensions of a container must be considered to dermine its volume.
> > :true
>
> A great way to figure out the volume of a trombone.
>
> > In 1977, the Earth revolved around two suns. : false
>
> With questions like this, an intelligence, artificial or not, might
> wonder why they are only up to 350000.
>
> --
> <J Q B>
--
Philip Jackson
McKinstry's response is typical of his previous responses, continuing to
make the clearly false claim that a sufficiently large pseudo-fact base
will somehow equate to (or test for) intelligence. McKinstry has yet to
answer any of the technical arguments against his approach which have
been posted by me and several others to this group. When pressed with
these technical arguments his response has been to reply angrily, both
in public and private correspondence, with language sometimes verging on
physical threats. If history remembers McKinstry, it will remember these
things about him.
In many ways it is a waste of time responding to his posts about MISTIC.
On the other hand, if we want to preserve comp.ai.* as a forum for (at
least some) reasoned discourse, then we must occasionally deny clearly
false claims by people who obviously know little about AI.
Phil Jackson
--------------------------------------------------------------------
"...for the word is the sole sign and the only certain mark of the
presence of thought hidden and wrapt up in the body..." -- Descartes
--------------------------------------------------------------------
Standard Disclaimers. <pjac...@ic.net>
Christopher McKinstry
John Nagle wrote:
> Well, the "Cyc" database is rather like that, although better
> organized for machine processing.
>
> John Nagle
Cyc is private, and not available to the public. As well, Cyc is top
down... it takes well trained knowledge engineers to enter data into
Cyc.
MISTIC is made by the public, for the public(it will be released after
we publish, of course). It is completely bottom up. Everything that
requires a knowledge engineer on Cyc, will be discovered automatically
with anaylsis of the MISTIC db.
MISTIC's primary purpose is to build a map of human consciousness. The
only people that have ever seen an actual visual map (adaptive-subspace
self-organizing map) of human consciousness are on the MISTIC team...
it's lowres right now, but we know what it looks like (very fractal)...
most everyone else will have to wait until after it makes the cover of
Science.
David G. Mitchell
In article <3337F1...@clickable.com>,
Christopher McKinstry <ch...@clickable.com> wrote:
>Jim, MISTIC has a very simple purpose... build a test that can tell the
>difference between a human and a random system with a very high degree
>of accracy.
Please explain;
- what you mean by a "random system",
- why distinguishing humans from such a system is an interesting task,
- what role your collection of "facts" has in performing this task.
David
Jim Balter
Christopher McKinstry wrote:
> This project has been set up to continue to collect data long after both
> you and I are dead. It's bigger than me, it's bigger than you, and more
> important than either of us.
What is big here is McKinstry's ego, like the ego of every crackpot
who thinks they have discovered or invented something of significance.
--
<J Q B>
Philip Jackson
Christopher McKinstry wrote:
>
> John Nagle wrote:
>
> > Well, the "Cyc" database is rather like that, although better
> > organized for machine processing.
> >
> > John Nagle
Both Cyc and MISTIC follow a similar approach of assembling a large fact
base as a route toward achieving human-level AI. This approach in itself
cannot be sufficient to achieve human-level AI -- babies are not born
knowing all the commonsense facts in Cyc, they acquire these facts as a
result of a general-purpose innate intelligence that has yet to be
emulated.
Beyond this surface level similarity, Cyc is quite different, and far
more advanced, than the ramblings of MISTIC. Cyc provides an ontology
for real-world knowledge within which one can state both commonsense
facts and their exceptions. Cyc supports inference and natural language
disambiguation. McKinstry has not yet posted any description of how
MISTIC might provide these things. Cyc is a valid large-scale
contribution to the field of AI. MISTIC is not.
>
> Cyc is private, and not available to the public.
Not exactly -- Cyc is commercially available, with limitations.
> As well, Cyc is top
> down... it takes well trained knowledge engineers to enter data into
> Cyc.
The entry of knowledge into Cyc by well-trained engineers has been both
a bottom up and top down process.
>
> MISTIC is made by the public, for the public(it will be released after
> we publish, of course).
Which means it is still private, though it has been created by public
contributions. McKinstry is like an inverted Robin Hood, taking from the
public to give to himself, though he has little chance of doing anything
useful with the input he is gathering.
> It is completely bottom up. Everything that
> requires a knowledge engineer on Cyc, will be discovered automatically
> with anaylsis of the MISTIC db.
How? What kind of analysis? This is a very challenging research topic.
McKinstry has not yet posted anything substantive about such things will
be discovered automatically -- Rather he has stated vaguely:
"My assertion is that [properties of human intelligence such as
creativity, reasoning, problem-solving, learning] are emergent
properties of a sufficiently large MIST database. Statistical systems,
such as neural nets can effectively cluster the raw MIST data and
generalize responses to stimuli that have not been seen by the system
before. Just as humans do."
This statement by him is completely ludicrous.
> MISTIC's primary purpose is to build a map of human consciousness.
Again, McKinstry tosses terms around loosely. To suggest that a
collection of true/false pseudo-facts is a map of either consciousness
or intelligence is ridiculous.
> The only people that have ever seen an actual visual map (adaptive-subspace
> self-organizing map) of human consciousness are on the MISTIC team...
No, these may be the only people who have deluded themselves into
thinking they are looking at a map of human consciousness.
> it's lowres right now, but we know what it looks like (very fractal)...
The reason McKinstry thinks it looks fractal is suggested by the
following excerpt of a previous posting:
CM: "For example, MIST items cluster reasonable well when simply sorted
alaphabetically."
PCJ: "No. For perhaps any two English sentences that you have sorted
together
alphabetically and claim cluster together well, one can probably easily
construct
other sentences that would fit between them and cluster differently. For
example, suppose MIST contained the following two sentences which sort
and cluster together:
Potatoes are good to eat. : true
Potatoes are nutritious. : true
"Then one could easily add the following sentences between them:
Potatoes are gooey : false
Potatoes are noisy : false"
Unlike the MISTIC pseudo-fact base, a truly fractal structure has a
definite pattern (which repeats at different scales of observation), and
which can be described mathematically and recreated exactly. In contrast
the MISTIC pseudo-fact base has a completely arbitrary structure that
can be manipulated ad nauseum to create any kind of pattern desired.
> most everyone else will have to wait until after it makes the cover of
> Science.
Don't hold your breath.
Christopher McKinstry
David G. Mitchell wrote:
> Please explain;
> - what you mean by a "random system",
all current non-human systems would be classified as random systems.
specifically a random system is a system that score .5 (half
correct/half incorrect) on a MIST Test.
> - why distinguishing humans from such a system is an interesting task,
because and deviation from random can be amplified. for example a
feedback system such as a neural net could be trained on MIST items.
with a large enough number of items, any "chance correctness" can be
amplified by that feedback system. thus, eventually in a MIST domain, it
would become increasingly hard to distinguish human from machine.
> - what role your collection of "facts" has in performing this task.
i started the project and maintain the systems to collect the data.
Christopher McKinstry
jim, just ignore the whole thing. take a deep breath. the data
collection can't possibly hurt you, so don't worry about it. take your
destructive tendencies elsewhere.
--
Christopher McKinstry
Philip Jackson wrote:
> On the other hand, if we want to preserve comp.ai.* as a forum for (at
> least some) reasoned discourse, then we must occasionally deny clearly
> false claims by people who obviously know little about AI.
i have made no false claims. notice no one with any actual AI experience
or letters after their name have objected the the concepts of the MISTIC
project. Simply because there is nothing wrong with collecting data.
It is impossible to predict what the results of training a neural net or
other system on MISTIC data would be(I have my hypothesis and I will
test it). I suggest you just wait and see, and keep your antiscience
attitude to yourself.
A little history: MISTIC was created after Marvin Minsky suggested to me
into private corospondence that a "very large corpus" of MIST items
would be useful in training a system to respond like a human. There was
just no way of collecting the data at the time. Now there is.
Philip Jackson
Christopher McKinstry wrote:
>
> Philip Jackson wrote:
>
> > On the other hand, if we want to preserve comp.ai.* as a forum for (at
> > least some) reasoned discourse, then we must occasionally deny clearly
> > false claims by people who obviously know little about AI.
>
> i have made no false claims.
Regarding properties of human intelligence such as reason, creativity,
problem-solving, and learning, you wrote:
"I don't have to build in these properties. My assertion is that they
are
emergent properties of a sufficiently large MIST database. Statistical
systems,
such as neural nets can effectively cluster the raw MIST data and
generalize
responses to stimuli that have not been seen by the system before. Just
as
humans do."
In my view this is a clearly false claim. It suggests that the
technology already exists for statistical systems / neural nets to
effectively cluster the raw data, generalize responses and create
emergent behavior duplicating human-level reasoning, creativity,
problem-solving and learning. In fact the opposite is true: There is no
existing computer-based statistical process or neural net algorithm that
can do this, as anyone who has actually published work on these topics
could confirm. And to suggest that a future computer-based, purely
statistical or neural net technology will do this without describing how
would also, in my view, be a misleading statement. Finally to suggest
that it is in principle possible because human intelligence is based on
neural nets would be vacuous and misleading: Neural net algorithms are a
simplification of human neural processing -- we do not yet understand
how the brain achieves human intelligence, and it may be that human
intelligence requires processes that are a hybrid of higher-level
symbolic processes and lower-level neural / associationist processes.
> notice no one with any actual AI experience
> or letters after their name have objected the the concepts of the MISTIC
> project.
Actually I have AI experience and publications. Perhaps many others have
not bothered to criticize MISTIC because your claims are too silly.
> Simply because there is nothing wrong with collecting data.
I have no objections to collecting data, though I think what you are
collecting is not very useful or meaningful data.
> It is impossible to predict what the results of training a neural net or
> other system on MISTIC data would be(I have my hypothesis and I will
> test it).
On the contrary, given knowledge of current neural net technology it is
quite valid to predict that the behavior you claim will emerge will not
occur. I have provided specific small examples and problems with your
approach, to which you have not responded substantively.
> I suggest you just wait and see, and keep your antiscience
> attitude to yourself.
My attitude is pro-science and what you are doing is not science: It is
pro-science of me to offer logical criticisms and problems for your
approach, and very anti-science of you to ignore these problems and
claim that human-level intelligence will magically emerge from your
collection of pseudo-facts.
> A little history: MISTIC was created after Marvin Minsky suggested to me
> into private corospondence that a "very large corpus" of MIST items
> would be useful in training a system to respond like a human. There was
> just no way of collecting the data at the time. Now there is.
Professor Minsky would have to speak for himself, of course. It is one
thing to suggest that a large set of commonsense facts could be useful
in developing or training an AI system, especially if there are no
constraints placed on the technology used within such a system.
However, I sincerely doubt that Dr. Minsky would endorse your claim that
human-level reason, creativity, problem-solving and learning could
emerge from neural nets clustering and generalizing your collection of
raw MIST data. I would guess that if he wrote something to you, you did
not understand what he wrote and are drawing many conclusions and making
many claims for MISTIC that he would not agree with, because you
apparently don't understand the technologies you are talking about.
SCN User
When my best friend Naiden and I played the game of kings as teen-
agers, Naiden laughed at my attempt to create a chess computer by
listing on sheets of paper all best responses to his chess moves.
He pointed out that I could never gather up a list of all gambits
and likewise people are now rushing to critique the MIST Project.
Thus Philip Jackson comments on what Christopher McKinstry wrote:
>> MISTIC is made by the public, for the public (it will be released
>> after we publish, of course).
> Which means it is still private, though it has been created by
> public contributions. McKinstry is like an inverted Robin Hood,
> taking from the public to give to himself, though he has little
> chance of doing anything useful with the input he is gathering.
Mr. McKinstry has the same chance that T. Edison had to invent an
electric light, or that U.S. President Jimmy Carter had to create
peace between Egypt and Israel. Philip Jackson has a clever wit,
with his remark above about Robin Hood, but the world needs those
who dream and fail, and then dream again as perhaps Mr. McKinstry
does in his AI work. Naiden laughed at Murray in 1962, but 1997:
/^^^^^^^^^^^\ Web-borne Mentifex (q.v.) AI Meme /^^^^^^^^^^^\
/visual memory\ semantic ________ / auditory \
| /--------|-------\ memory / syntax \ | speech memory |
| | recog-|nition | \________/<--|-------------\ |
| ___|___ | | flush-vector| spiral| _______ | |
| /image \ | __|___ ___V___ loop| /stored \ | |
| / percept \ | /deep \<-----/lexical\<---|--/ phonemes\| |
| \ engrams /<--|-->/concepts\--->/concepts \---|->\ of words/ |
| \_______/ | \________/ \_________/ | \_______/ |
Footnote to history: Naiden's father, who taught at The Lakeside
School in Seattle, caused Microsoft to come into existence by re-
questing of the Lakeside Mothers Club that they give one thousand
dollars for the purpose of buying computer timeshare for students
with the result that Gates and Allen used up all the time and....
For photo of Naiden see NYT 25.MAR.1993 C-19, facing John Hersey.
Daniel D Raasch
As an undergrad student who's planning on majoring in AI, I just
have a couple of questions regarding the MIST project:
1) Wouldn't most people here agree that McKinstry's Minimum
Intelligent Signal Test is a real leap beyond the Turing Test?
It eliminates a great amount of subjectivity, provides quantifiable data,
and effectively creates a scale of intelligence for any given test. The
fact that you could administer a MIST to a chimp, child, or computer seems
to me to be a wonderful step towards pinning down the elusive quality of
intelligence. We may not be close to achieving an algorithm that could
pass it, but we weren't when Turing proposed his idea, either.
So if someone disagrees with the idea that McKinstry's MIST (the
test alone, not the corpus) is a better idea then what we've had previously,
would they speak up and say why?
2) I'll admit that the details of how McKinstry purports to
alchemize the data in the corpus into an algorithm that embodies
representational knowledge seem a little vague to me. Maybe Mr. McKinstry
would explain in a little more detailed manner how he expects a
statistical or neural net based system to evaluate the individual entries?
I have special interest in statements that are equally true and false,
such as, "Humans are sometimes blue."
Thanks. I appreciate it.
-Daniel DW Raasch
University of Minnesota
Chris Hooley
>
>The MIST Item Corpus Project is the largest (in terms of participants)
>Artificial Intelligence project in human history. This is due entirely
>to the WWW and people like you. The goal of this project is to make a
>high resolution conceptual map of human cognition.
Amazing that scientists would ever resist learning something new, eh?
I think you'll demonstrate a close mimic of human mentation and teach
us all a lot besides.
Weighting is an issue, but if repeated inputs count repeatedly it
should work out.
Somehow the core stuff, like desire to continue living and to
replicate have to settle deep in the db. Intention makes choices
count, seperates the impersonal man in the Chinese room from the
vulnerable self. If your system self-organizes as though it had
something to lose, it'll shock the naysayers.
Along those lines, how are you recognizing statements of organization?
Many of our most important meanings are rules about organizing
meanings. In other words, the hierarchy of meaning isn't democratic.
The only other thing is that your mind will be affected by the medium
of collection. It will most resemble an adolescent boy. :-)
Good science!
Chris
Philip Jackson
SCN User (presumably Arthur T. Murray) wrote:
>
> When my best friend Naiden and I played the game of kings as teen-
> agers, Naiden laughed at my attempt to create a chess computer by
> listing on sheets of paper all best responses to his chess moves.
> He pointed out that I could never gather up a list of all gambits
Naiden's criticism of your approach to chess was valid.
> and likewise people are now rushing to critique the MIST Project.
I wouldn't say people are "rushing" to critique MIST. There are a few
people who occasionally respond to McKinstry's repeated overblown
claims.
> Thus Philip Jackson comments on what Christopher McKinstry wrote:
>
> >> MISTIC is made by the public, for the public (it will be released
> >> after we publish, of course).
>
> > Which means it is still private, though it has been created by
> > public contributions. McKinstry is like an inverted Robin Hood,
> > taking from the public to give to himself, though he has little
> > chance of doing anything useful with the input he is gathering.
>
> Mr. McKinstry has the same chance that T. Edison had to invent an
> electric light, or that U.S. President Jimmy Carter had to create
> peace between Egypt and Israel.
No, McKinstry has far less chance than Thomas Edison. Edison invented
the electric light through a patient search for an adequate filament
based on his theory of how an electric light would work. Edison had a
good chance of succeeding because his theory was correct, and he had the
ability to try many different filaments and the will to persevere.
McKinstry on the other hand does not have an adequate theory of how to
make use of the data he is gathering. Rather he claims that human
intelligence can emerge from a neural net performing simple correlations
and clustering on his collection of pseudo-facts. Such claims
demonstrate he does not understand the limitations of the technologies
he is talking about.
Perhaps President Carter had a theory of how to achieve peace between
Egypt and Israel. Diplomacy may also be a trial and error process
requiring great patience. Beyond that, there may be little comparison
between one's chances for solving a difficult but precise technical
problem, and one's chances for solving a difficult, imprecise
geopolitical problem.
> Philip Jackson has a clever wit,
> with his remark above about Robin Hood, but the world needs those
> who dream and fail, and then dream again as perhaps Mr. McKinstry
> does in his AI work.
The world needs inspiration, but it also needs concrete work: Edison
said that genius is 1% inspiration and 99% perspiration. The
perspiration is the work of developing the details for a workable theory
that will support the inspiration, and the work of conducting the
experiments to support or disprove the theory.
So far, McKinstry appears not to have done the necessary concrete work
to achieve the results he claims will happen. From that perspective, he
does not seem to have very much inspiration, either. However, I would
encourage him to take my advice to heart, rather than rejecting and
ignoring it: If he really studies the problems I have posed, and the
published work on neural nets and other AI technologies, perhaps he can
modify his approach and achieve something worthwhile.
> Naiden laughed at Murray in 1962, but 1997:
>
> /^^^^^^^^^^^\ Web-borne Mentifex (q.v.) AI Meme /^^^^^^^^^^^\
> /visual memory\ semantic ________ / auditory \
> | /--------|-------\ memory / syntax \ | speech memory |
> | | recog-|nition | \________/<--|-------------\ |
> | ___|___ | | flush-vector| spiral| _______ | |
> | /image \ | __|___ ___V___ loop| /stored \ | |
> | / percept \ | /deep \<-----/lexical\<---|--/ phonemes\| |
> | \ engrams /<--|-->/concepts\--->/concepts \---|->\ of words/ |
> | \_______/ | \________/ \_________/ | \_______/ |
>
Perhaps Naiden would still laugh, and still be right. However, I will
keep an open mind and put it on my list to read your writings about
Mentifex and try to give some feedback.
> Footnote to history: Naiden's father, who taught at The Lakeside
> School in Seattle, caused Microsoft to come into existence by re-
> questing of the Lakeside Mothers Club that they give one thousand
> dollars for the purpose of buying computer timeshare for students
> with the result that Gates and Allen used up all the time and....
Perhaps by helping to bring Gates and Allen together, and giving them
some programming experience and knowledge of computers, this was one
important causative factor in the eventual creation of Microsoft, though
they of course would be the ones to say how important it was.
Relative to dreams and work, it took great vision and much hard work for
them to leave Harvard and start Microsoft when the Altair appeared in
the 1970's. However, Microsoft might still have been a failure without
great business acumen, business luck, and continued hard work later on.
Christopher McKinstry
Philip Jackson wrote:
<snip>
> "My assertion is that [properties of human intelligence such as
> creativity, reasoning, problem-solving, learning] are emergent
> properties of a sufficiently large MIST database. Statistical systems,
> such as neural nets can effectively cluster the raw MIST data and
> generalize responses to stimuli that have not been seen by the system
> before. Just as humans do."
that's my assertion, and i'm sticking to it.
> CM: "For example, MIST items cluster reasonable well when simply sorted
> alaphabetically."
this is quite out of context, but it happens to be true. however, MIST
self organize much, much better. the MISTIC sampling captures what we
are at very high resolution. everything we are can be derived from a
sufficiently large MIST item database. such a database needs no exteranl
references to the world, no knowledge engineers.
one could change every word in the database to something random, and
looking at the self organized clusters, one could very easily translate
it back to english, because it's a map of thought not language.
MISTIC makes the conceptual structures of our consciousness just as
visible as a cat scan or mri makes visible the structures of our brains.
the concept, process and the math are identical. the only difference is
that human conscious is an object that exists in a much higher
dimensionality, than the 3-d physical brain.
MISTIC is high dimensional tomography.
CYC is an attempt to build a high dimensionality object from scratch,
without ever seeing one.
Christopher McKinstry
Chris Hooley wrote:
> The only other thing is that your mind will be affected by the medium
> of collection. It will most resemble an adolescent boy. :-)
there's a whole pile of items in the db like the following:
do you feel lonely at time? true
when peole laugh at you do you sometime feel bad? true
the older you get the more mortal you feel. true
and so on... thus, the human emotional factor is also captured in
MISTIC. there's some very advanced stuff in there. it is becomming very
human. a complete adualt consciousness can be captured with enough MIST
items... not just an adolescent boy. (why you specified a sex, i have no
idea).
Christopher McKinstry
Philip Jackson wrote:
> No, McKinstry has far less chance than Thomas Edison. Edison invented
> the electric light through a patient search for an adequate filament
> based on his theory of how an electric light would work. Edison had a
> good chance of succeeding because his theory was correct, and he had the
> ability to try many different filaments and the will to persevere.
it does not matter how you extract the meaning form MISTIC. my assertion
is that it will be captured. in the very worst case, you have a db with
all possible items... or at least more items that a resonable person can
generate in a life time (which is possible). it such a case, you could
not tell the db from a person in a human lifetime using nothing more
sophisticated that linear patter matching.
however, i think it is much more reasonable to think that we will
cluster and analyze the MISTIC data so that generalization can be made
and the db will need nowhere near the amount of data specified above.
MISTIC is the minimum way a human order consciousness can be specified.
it's just digitization.
if i were to take a random sample for a CD(compact disk),
(position,value) i would not get much meaning... but if i took a great
random deal of samples, i would get a lower resolution version of that
CD. i could make reasonable predictions on the value of any position,
simply by knowing it's position only. the more data i had, the more
likely that my estimate will be correct.
consciousness is like a CD. it is structured. a CD is just a very
complex 1-d object. consciousness as captured in MIST items is just an
exercise in adding dimensions.
Christopher McKinstry
Daniel D Raasch wrote:
>
>
> As an undergrad student who's planning on majoring in AI, I just
> have a couple of questions regarding the MIST project:
>
> 1) Wouldn't most people here agree that McKinstry's Minimum
> Intelligent Signal Test is a real leap beyond the Turing Test?
> It eliminates a great amount of subjectivity, provides quantifiable data,
> and effectively creates a scale of intelligence for any given test. The
> fact that you could administer a MIST to a chimp, child, or computer seems
> to me to be a wonderful step towards pinning down the elusive quality of
> intelligence. We may not be close to achieving an algorithm that could
> pass it, but we weren't when Turing proposed his idea, either.
>
i don't think anyone disagrees with this.
> So if someone disagrees with the idea that McKinstry's MIST (the
> test alone, not the corpus) is a better idea then what we've had previously,
> would they speak up and say why?
>
> 2) I'll admit that the details of how McKinstry purports to
> alchemize the data in the corpus into an algorithm that embodies
> representational knowledge seem a little vague to me. Maybe Mr. McKinstry
> would explain in a little more detailed manner how he expects a
> statistical or neural net based system to evaluate the individual entries?
> I have special interest in statements that are equally true and false,
> such as, "Humans are sometimes blue."
intelligence can't be inserted into a system; it must be developed
through the system's interaction with the outside environment.
the key is to build a tightly coupled feedback loop. MISTIC allows for a
system (i don't know what type, which is what bothers everyone) to
feedback againt what in effect is a binary model of the human experience
of reality.
build it, and it will come.
Christopher McKinstry
Daniel D Raasch wrote:
> I have special interest in statements that are equally true and false,
> such as, "Humans are sometimes blue."
statements that areequlayy true and false have not value in a MIST test,
as it doesn't matter how you answer them. a coin can answer them just as
well as a person.
the important items are those which require actual human consciousness
to answer. consensus knowledge. these are what will allow us to measure
and judge systems againts the human standard. once we can measure
something on a continuous scale of fitness, we can gradually imporve it,
and eventually duplicate it.
Philip Jackson
Christopher McKinstry wrote:
>
> Philip Jackson wrote:
> <snip>
> > "My assertion is that [properties of human intelligence such as
> > creativity, reasoning, problem-solving, learning] are emergent
> > properties of a sufficiently large MIST database. Statistical systems,
> > such as neural nets can effectively cluster the raw MIST data and
> > generalize responses to stimuli that have not been seen by the system
> > before. Just as humans do."
>
> that's my assertion, and i'm sticking to it.
Just to be clear for other readers, the statements in quotes above are
McKinstry's, not mine. He is repeating above that these statements in
quotes are his assertion. His <snip> elided my reference that he had
made the quoted statements. In my view, McKinstry's statements are a
very overblown, ludicrous claim and he has not yet provided any
substantiation for his claim.
[...]
> everything we are can be derived from a
> sufficiently large MIST item database. such a database needs no exteranl
> references to the world, no knowledge engineers.
Yes, and pigs can fly.
> one could change every word in the database to something random, and
> looking at the self organized clusters, one could very easily translate
> it back to english, because it's a map of thought not language.
>
> MISTIC makes the conceptual structures of our consciousness just as
> visible as a cat scan or mri makes visible the structures of our brains.
> the concept, process and the math are identical. the only difference is
> that human conscious is an object that exists in a much higher
> dimensionality, than the 3-d physical brain. [...]
You have yet to prove or substantiate any of your claims. Put up or shut
up. A claim that all this will happen eventually, without any
substantive description of how, is just as vacuous and unsubstantiated
as all your other claims.
Philip Jackson
Christopher McKinstry wrote:
>
> Daniel D Raasch wrote:
> >
> >
> > As an undergrad student who's planning on majoring in AI, I just
> > have a couple of questions regarding the MIST project:
> >
> > 1) Wouldn't most people here agree that McKinstry's Minimum
> > Intelligent Signal Test is a real leap beyond the Turing Test?
No, it's a great leap backward. Your asking this question shows how
important it is for AI researchers to pay attention to and debunk
McKinstry, because his claims distract from the real accomplishments
that have been made toward AI, and in particular distract from the
contribution that Turing made in proposing his test. Turing must be
rolling over in his grave.
Perhaps the most important advantage that the Turing Test has over MIST
in testing for intelligence is that the Turing Test is dynamic and
interactive, enabling one intelligence to engage others and test and
compare the ability of the others to reason, solve problems, be
creative, and to learn dynamically.
In contrast, MIST is at any point in time a set of true/false labeled
pseudo-facts: It might be used to test rote recall but that is clearly
not the same as testing for intelligence. McKinstry claims that a neural
net system could achieve human-level intelligence by being trained with
the MIST database, though his claims are not substantiated, as he in
effect admits below. Even if this were possible, one could not use the
MIST database to test the dynamic reasoning and problem-solving
intelligence of such a system. One would have to use the Turing Test or
something very like it, to test for such dynamic intelligence qualities.
> > It eliminates a great amount of subjectivity, provides quantifiable data,
> > and effectively creates a scale of intelligence for any given test.
Sorry, I have to disagree. It appears that the MIST collection of
true/false pseudo-facts will retain substantial subjectivity, that it
will not provide any valid quantifiable data, that it will not
effectively enable testing for intelligence.
> > The
> > fact that you could administer a MIST to a chimp, child, or computer seems
> > to me to be a wonderful step towards pinning down the elusive quality of
> > intelligence. We may not be close to achieving an algorithm that could
> > pass it, but we weren't when Turing proposed his idea, either.
> >
>
> i don't think anyone disagrees with this.
Which shows how little McKinstry understands the criticisms that have
been made for MIST. I've written that the MIST approach has no chance of
capturing the "elusive qualities" of intelligence such as reason,
creativity, problem-solving and learning. Implicit (and sometimes
explicit) in my criticisms has been that the MIST approach has little
chance of testing for these elusive qualities of intelligence, either.
Others besides me have made similar criticisms.
>
> > So if someone disagrees with the idea that McKinstry's MIST (the
> > test alone, not the corpus) is a better idea then what we've had previously,
> > would they speak up and say why?
Yes, I've spoken up for some time. In addition to the reasons given
above, I wrote in July 1996, in response to McKinstry's "open request"
for feedback on comp.ai.*:
"I have a problem with the theory behind the project. Namely, you do not
seem to have any provision in your architecture for the system to be
creative, to reason, to solve problems, to learn, to invent questions,
etc. All you seem to have is a catalog of true/false statements and a
claim that if you have enough such statements, a system can be trained
from them to appear to be equivalent to a human being answering yes/no
questions. You have not specified the training mechanism except to
suggest that "statistical correlation" and "neural nets" could be used
-- yet these techniques are inadequate to the task."
I also wrote: ".. even very precise English statements can have
alternate, context-dependent interpretations in which they may be
considered either true or false. This is another flaw in the MIST
approach, it does not support context-dependent reasoning and
interpretation. Again, the point is that MIST would fail to answer even
easier questions, that most humans would be able to answer successfully
via reasoning."
And I wrote "[MIST makes] the problem of mimicing human knowledge more
difficult by restricting the knowledge [...] to English assertions that
can be labeled "true" and "false". A truly intelligent system must be
able to reason about assertions that are uncertain or context
dependent."
And I wrote: "A human being could pose a yes/no question about whether a
certain answer is the solution to a certain problem. Another human being
could answer this yes/no question, using reason and problem solving
skills to determine the correct solution to the problem. This is not
something that MIST would be able to do, even though it involves yes/no
questions."
> >
> > 2) I'll admit that the details of how McKinstry purports to
> > alchemize the data in the corpus into an algorithm that embodies
> > representational knowledge seem a little vague to me.
Yes, McKinstry has not provided such details, and now, thanks to your
question, admits below that he cannot.
> > Maybe Mr. McKinstry
> > would explain in a little more detailed manner how he expects a
> > statistical or neural net based system to evaluate the individual entries?
I have also asked how such a system could demonstrate reasoning ability
when presented with statements and questions that are not in the
existing database -- This is a real problem, which McKinstry has ignored
or blithely promised that statistical systems can solve via
"generalization".
> > I have special interest in statements that are equally true and false,
> > such as, "Humans are sometimes blue."
Right. This is a big flaw and weakness in McKinstry's approach, which I
have previously pointed out.
>
> intelligence can't be inserted into a system; it must be developed
> through the system's interaction with the outside environment.
>
> the key is to build a tightly coupled feedback loop. MISTIC allows for a
> system (i don't know what type, which is what bothers everyone) to
> feedback againt what in effect is a binary model of the human experience
> of reality.
Here, McKinstry blithely admits he doesn't know how an intelligent
system could be trained with his MIST pseudo-data. In effect, he admits
he cannot support his claim that a neural net system could be trained
with the MIST pseudo-fact base to achieve human-level intelligence
reasoning, problem-solving, creativity and learning.
>
> build it, and it will come.
There are good technical reasons to believe that no intelligent system
will ever be used in conjunction with MIST. I've already noted that MIST
potentially contains lots of nonsense and subjective pseudo-facts
(perhaps screened only by McKinstry), which limits its usefulness.
Another problem is that from a training perspective, an intelligent
system would benefit from an incremental growth path from (carefully
screened, non-subjective) simple knowledge to more complex knowledge,
rather than by trying to generalize over a large base of subjective
nonsense and pseudo-facts.
If and when a human-level AI is constructed, it is much more likely that
it may read through the MIST collection of pseudo-facts and itself add
the statement:
MIST is stupid: True.
Jim Balter
Christopher McKinstry wrote:
>
> Daniel D Raasch wrote:
> >
> >
> > As an undergrad student who's planning on majoring in AI, I just
> > have a couple of questions regarding the MIST project:
> >
> > 1) Wouldn't most people here agree that McKinstry's Minimum
> > Intelligent Signal Test is a real leap beyond the Turing Test?
> > fact that you could administer a MIST to a chimp, child, or computer seems
> > to me to be a wonderful step towards pinning down the elusive quality of
> > intelligence. We may not be close to achieving an algorithm that could
> > pass it, but we weren't when Turing proposed his idea, either.
> >
>
> i don't think anyone disagrees with this.
Which shows what your thinking is worth.
--
<J Q B>
Jim Balter
Philip Jackson wrote:
> MIST is stupid: True.
Let's reiterate that. And let's add:
Chris McKinstry is an ego-ridden blowhard: true
For those who think that MIST is viable, let them consider the
criteria for allowing statements into the MIST database.
Let them consider what *validation* mechanism is used to guarantee
that the content is *intelligent* rather than *stupid*.
Let them consider the consequence of contributors willfully
submitting statements that they do not believe are true. By, say,
a random process.
--
<J Q B>
Jim Balter
Christopher McKinstry wrote:
>
> Daniel D Raasch wrote:
>
> > such as, "Humans are sometimes blue."
>
> as it doesn't matter how you answer them. a coin can answer them just as
> well as a person.
Random answers to "Are humans are sometimes blue?" will not suffice.
Which is one of many examples of how the Turing Test is related to
intelligence and MISTIC is not.
--
<J Q B>
Jim Balter
Christopher McKinstry wrote:
>
> Chris Hooley wrote:
> > The only other thing is that your mind will be affected by the medium
> > of collection. It will most resemble an adolescent boy. :-)
>
> there's a whole pile of items in the db like the following:
>
> do you feel lonely at time? true
> when peole laugh at you do you sometime feel bad? true
> the older you get the more mortal you feel. true
Sounds like a Scientology questionnaire.
Actually I think MIST is an intelligence test of sorts.
Intelligent people understand what's wrong with it.
Sometimes I feel lonely. True.
Sometimes I don't feel lonely. True.
Sometimes I think I'd be happy with just another dollar. True.
Sometimes I dont't think I'd be happy with just another dollar. True.
Sometimes I think I'd be happy with just another 2 dollars. True.
Sometimes I don't think I'd be happy with just another 2 dollars. True.
Sometimes I think I'd be happy with just another 3 dollars. True.
Sometimes I don't think I'd be happy with just another 3 dollars. True.
Sometimes I think I'd be happy with just another 4 dollars. True.
Sometimes I don't think I'd be happy with just another 4 dollars. True.
Sometimes I think I'd be happy with just another 5 dollars. True.
Sometimes I don't think I'd be happy with just another 5 dollars. True.
Sometimes I think I'd be happy with just another 6 dollars. True.
Sometimes I don't think I'd be happy with just another 6 dollars. True.
Sometimes I think I'd be happy with just another 7 dollars. True.
Sometimes I don't think I'd be happy with just another 7 dollars. True.
Sometimes I think I'd be happy with just another 8 dollars. True.
Sometimes I don't think I'd be happy with just another 8 dollars. True.
Sometimes I think I'd be happy with just another 9 dollars. True.
Sometimes I don't think I'd be happy with just another 9 dollars. True.
Sometimes I feel just like a monkey at a typewriteglumphghx. Yep.
> Good science!
> Chris
You flunk.
--
<J Q B>
Philip Jackson
Jim Balter wrote:
>
> Philip Jackson wrote:
>
> > MIST is stupid: True.
>
> Let's reiterate that. And let's add:
>
> Chris McKinstry is an ego-ridden blowhard: true
McKinstry's inability to respond to the technical problems that have
been pointed out for his approach, suggests his ego is overriding
whatever intelligence he possesses.
Cheers,
Philip Jackson
Christopher McKinstry wrote:
>
> Chris Hooley wrote:
> > The only other thing is that your mind will be affected by the medium
> > of collection. It will most resemble an adolescent boy. :-)
>
> there's a whole pile of items in the db like the following:
>
> do you feel lonely at time? true
> when peole laugh at you do you sometime feel bad? true
> the older you get the more mortal you feel. true
>
> MISTIC. there's some very advanced stuff in there. it is becomming very
> human. a complete adualt consciousness can be captured with enough MIST
> items... not just an adolescent boy. (why you specified a sex, i have no
> idea).
Here McKinstry forgets another element of human intelligence and
consciousness that is context dependent. (Perhaps because he has no idea
what sex is? Perhaps McKinstry is really a an 8 year old?) MISTIC could
flunk the first Turing Test it gets when someone asks whether it's male
or female, and then asks further questions specific to adult males and
females.
Philip Jackson
Christopher McKinstry wrote:
>
> Daniel D Raasch wrote:
>
> > I have special interest in statements that are equally true and false,
> > such as, "Humans are sometimes blue."
>
> statements that areequlayy true and false have not value in a MIST test,
> as it doesn't matter how you answer them. a coin can answer them just as
> well as a person.
No, a coin cannot. Such statements are context dependent (a concept
McKinstry apparently doesn't understand, or else boneheadedly refuses to
address), and a person can answer them differently depending on the
context of a conversation. To answer such questions validly in a given
context requires human intelligence, and can't be done by coin flips
just as well as a person.
> the important items are those which require actual human consciousness
> to answer.
Which again include context dependent items.
> consensus knowledge. these are what will allow us to measure
> and judge systems againts the human standard. once we can measure
> something on a continuous scale of fitness, we can gradually imporve it,
> and eventually duplicate it.
More pie in the sky from McKinstry. Based on the false assumption that
MIST's hodgepodge of pseudo-facts and nonsense can be used to judge
systems against a "human standard" of intelligence, he suggests further
that such systems can be gradually improved to duplicate human
intelligence -- though he now admits he doesn't know how.
Philip Jackson
Christopher McKinstry wrote:
>
> Philip Jackson wrote:
> > the electric light through a patient search for an adequate filament
> > based on his theory of how an electric light would work. Edison had a
> > good chance of succeeding because his theory was correct, and he had the
> > ability to try many different filaments and the will to persevere.
>
> is that it will be captured. in the very worst case, you have a db
> with all possible items...
The set of all possible English sentences which might be labelled
true/false is in principle countably infinite. It is not possible for
MIST to capture all such sentences.
> or at least more items that a resonable person can
> generate in a life time (which is possible). it such a case, you could
> not tell the db from a person in a human lifetime using nothing more
> sophisticated that linear patter matching.
> however, i think it is much more reasonable to think that we will
> cluster and analyze the MISTIC data so that generalization can be made
> and the db will need nowhere near the amount of data specified above.
So far as I know you have never really "specified" how much data you
might need. However, with 350,000 items of nonsense and pseudo-facts you
already have far more items than a reasonable person would want.
Moreover, I've previously given simple examples that show if one has the
ability to search and sort your items then a reasonable person could
easily find gaps and generate potentially meaningful sentences that do
not exist in the MIST nonsense-base. In general, evaluation of such
sentences would be context-dependent and require inference and
problem-solving, i.e. such sentences could not be evaluated by linear
pattern matching on the MIST nonsense-base. Nor could such sentences be
evaluated by simple clustering or correlation or existing neural net
algorithms trained on the MIST nonsense-base. A reasonable person could
easily distinguish the MIST nonsense-base (or a MISTIC system trained
with these methods) from a human intelligence. You now clearly admit you
don't know what kind of system and training procedure might attain
human-level intelligence from the MIST nonsense-base, and I've given
technical reasons why no one building or training an intelligent system
would want to make use of MIST.
So, my statement remains: You do not have a valid theory to guide your
efforts. Your chances of success are paltry, dismal. So far as I know,
you have not received any public endorsement of your efforts from anyone
who has published AI research in a recognized journal.
> MISTIC is the minimum way a human order consciousness can be specified.
> it's just digitization.
>
> if i were to take a random sample for a CD(compact disk),
> (position,value) i would not get much meaning... but if i took a great
> random deal of samples, i would get a lower resolution version of that
> CD. i could make reasonable predictions on the value of any position,
> simply by knowing it's position only. the more data i had, the more
> likely that my estimate will be correct.
>
> consciousness is like a CD. it is structured. a CD is just a very
> complex 1-d object. consciousness as captured in MIST items is just an
> exercise in adding dimensions.
Perhaps your consciousness and intelligence is a static,
context-independent set of nonsense and pseudo-facts -- it certainly
seems to be from your repetition of invalid claims. However, in general
human-level consciousness and intelligence are much more powerful. The
sampling approach you describe is far too weak and limited to capture
human consciousness and intelligence from any heap of nonsense, no
matter how deeply it is piled.
Bill House
Philip Jackson <pjac...@mail.ic.net> wrote in article
<333DA9...@mail.ic.net>...
>
> [snip]
>
> > > It eliminates a great amount of subjectivity, provides quantifiable
data,
> > > and effectively creates a scale of intelligence for any given test.
>
> Sorry, I have to disagree. It appears that the MIST collection of
> true/false pseudo-facts will retain substantial subjectivity, that it
> will not provide any valid quantifiable data, that it will not
> effectively enable testing for intelligence.
>
Although I have to agree that a database of factoids is not going to
automatically produce a human-level intelligence in a NN, I have to wonder
how much the validity of the factoids really matters. In the case of
humans, it's pretty obvious that people can believe the most outrageous
fantasies to be factual, but as long as they manage to survive long enough
to produce and raise progeny, they are able to persist. Untruth gets
dressed up as culture and tradition and life somehow goes on. If some kind
of true AI was given the MISTIC database as an input, the real test would
be how long it took for it to recognize the invalid factoids the database
contains. Perhaps good BS-detection is a component of intelligence?
Bill House
--
bho...@dazsi.com
http://www.dazsi.com
Ansel Morgan
Philip Jackson wrote:
> You have yet to prove or substantiate any of your claims. Put up or shut
> up. A claim that all this will happen eventually, without any
> substantive description of how, is just as vacuous and unsubstantiated
> as all your other claims.
what i claim is tangental to the project of data collection. don't worry
about it. just bite your tounge, ignore my claims and put some data in
the MISTIC database.
it doesn't matter if my claims turn out to be true or not. the data is
important.
Ansel Morgan
Jim, i will no longer respond to either you or phil jackson on this
matter. you two have spewed garbage on the comp.ai groups for years now,
and have never produced a single thing.
neither of you are qualified to comment.
as for my ego being big, damn right.
- how many tv interviews have you done? i've done well over 100.
- when was the last time your were on CNN? january 8, 1997 for me.
- how old were you when you made you first million? i was 17.
- when was the last time you were asked to be keynote speaker at a
convocation? february 28th, 1997 for me.
- do you drive a jaguar? i do.
remember that. i do, you (and phil) spew.
Jim Balter wrote:
>
> Philip Jackson wrote:
>
> > MIST is stupid: True.
>
> Let's reiterate that. And let's add:
>
> Chris McKinstry is an ego-ridden blowhard: true
>
> For those who think that MIST is viable, let them consider the
> criteria for allowing statements into the MIST database.
> Let them consider what *validation* mechanism is used to guarantee
> that the content is *intelligent* rather than *stupid*.
> Let them consider the consequence of contributors willfully
> submitting statements that they do not believe are true. By, say,
> a random process.
>
> --
> <J Q B>
--
Norman R. Gall
In article <333ECD...@clickable.com>, Ansel Morgan
<in...@clickable.com> wrote:
>Jim, i will no longer respond to either you or phil jackson on this
>matter.
Well, of course not. Phil and Jim have challenged your unsubstantiated
cliams with facts and arguments. You have responded by repeating the same
unsubstantiated claims like a mantra and are now sitting there wondering
why everyone else hasn't been mesmerised by the doctrinaire position you
keep repeating.
>you two have spewed garbage on the comp.ai groups for years now,
>and have never produced a single thing.
Sure they have. They simply haven't relied upon mass media and Usenet
newsgroups to be their measure of achievement.
>neither of you are qualified to comment.
I'd submit that you aren't qualified to make the claims that they comment
upon. Interestingly, no one else does either.
>as for my ego being big, damn right.
>- how many tv interviews have you done? i've done well over 100.
So has Manson, Madonna, and Camile Paglia - meaningless.
>- when was the last time your were on CNN? january 8, 1997 for me.
Seeing as you were talking about CRap6, this too is meaningless.
>- how old were you when you made you first million? i was 17.
Again, since we are talking about the plausibility of your claims, this is
meaningless as well.
>- when was the last time you were asked to be keynote speaker at a
>convocation? february 28th, 1997 for me.
I could make a snotty remark here, but I won't. Bruce Cockburn was the
keynote at my wife's convocation, but that doesn't qualify him to make
comments on, say, quantum chemistry - or *anything* for that matter.
>- do you drive a jaguar? i do.
Again, utterly irrelevant.
>remember that. i do, you (and phil) spew.
Your stream of red herrings, improper appeals to popularity and authority,
and straw persons are impressive, but simply fallacious.
Your problem remains the same as it has always been: you are
undereducated, mentally tortured, and so depressingly egomaniacally
self-absorbed that you cannnot see that whatever rationality you might
have had has left you.
Your responses are irrelevant. The 'theory' underyling your magnum opus AI
project is as thin as gossamer. And your infantile bawling regarding your
'accomplishments' are downright laughable.
If this is, in fact, your last squeal here regarding MIST, it is quite
approprate that it have nothing to do with the charges against you - since
MIST has nothing to do with AI.
Norm Gall
--
"All philosophy can do is to destroy idols. And that means not
creating a new one - for instance as in the 'absence of an idol.'
- L. Wittgenstein
Norman R. Gall
>K. Christopher McKinstry
>Homepage: Http://www.clickable.com/employees/chris
>World's Largest AI Project: Http://www.clickable.com/mist_corpus.html
>World's Coolest Interactive Drama http://www.cr6.com
And, by the way, 'Ansel'. You forgot to change the settings back in Netscape.
Ansel Morgan
My fault. I was supoed to change it back. Maybe were the same guy, right
Chris??
He said "yeah, what ever you say ans. just make sure when you quote me
from now on, you don't use caps. i don't like caps. too arogant."
Too arogant! I stopped believeing it was possible when I met you.
Chris said "make sure you delete my signature, change the identity back
to me, and get your own damn account."
To which I say: okay, okay, pay me money and I will.
Jim Balter
> what i claim is tangental to the project of data collection. don't worry
> about it. just bite your tounge, ignore my claims and put some data in
> the MISTIC database.
>
> it doesn't matter if my claims turn out to be true or not. the data is
> important.
"What can it hurt to call?" -- Dionne Wartlick
--
<J Q B>
David Hatch
Chris-
The "higher dimensionality" you refer to is associated with the square
root of minus
one ( i ) - the true symbol of consciousness. It has an unsolvable quality
to it,
which leads me to believe that project MYSTIC will come to a dead end. You
are
welcome to add this question to the database: Does the square root of minus
one have an easy solution? : false
See http://home.inreach.com/davhatch/Theory/index.html
for more about the identity between consciousness and the square root of
minus one ( i ).
- Dave
Christopher McKinstry <ch...@clickable.com> wrote in article
<333C59...@clickable.com>...
><snip>
> that human conscious is an object that exists in a much higher
> dimensionality, than the 3-d physical brain.
>
Jim Balter
Chris McKinstry, posing as Ansel Morgan wrote:
> as for my ego being big, damn right.
> - how many tv interviews have you done? i've done well over 100.
> - when was the last time your were on CNN? january 8, 1997 for me.
> - how old were you when you made you first million? i was 17.
> - when was the last time you were asked to be keynote speaker at a
> convocation? february 28th, 1997 for me.
> - do you drive a jaguar? i do.
Your mind is even more puny than I had imagined.
--
<J Q B>
gordon r. grieder
Norman R. Gall wrote:
> And, by the way, 'Ansel'. You forgot to change the settings back in Netscape.
LOL, I cought that too Norm, but you beat me to it.
(Chris' own mix of Ansel Adams and J.P. Morgan perhaps?)
--
gordon grieder caprice dist. ltd.
<mailto:go...@caprice.mb.ca> pgp key at <http://www.Four11.com>
K. Christopher McKinstry
Philip Jackson wrote:
> No, a coin cannot. Such statements are context dependent (a concept
> McKinstry apparently doesn't understand, or else boneheadedly refuses to
> address), and a person can answer them differently depending on the
> context of a conversation. To answer such questions validly in a given
> context requires human intelligence, and can't be done by coin flips
> just as well as a person.
phil, i'm not going to respond to you in public ever again.
however, if you had read the MISTIC materials, which it is obvious you
haven't, you would know that items such as "are you male?", are not
answered in a consistent fashion by most people and thus are not
permitted in a MIST test. Such questions were quite specifically
excluded from the very beginning.
MIST is about the MINIMUM Intelligent Signal. A single binary test in
complete isolation. MIST is the smallest test you can perform on an
intelligent entity or an entity proporting to be intelligent. "Are you
male?" is not, and cannot be a MIST item. on it's own, independent of
any other questions, it tells you nothing about the consciousness of the
entity being questioned. the entity could be a coin, and you wouldn't
know the difference.
A MIST item must be stable, first and formost. That means that it must
be answered, in isolation in a consistent fashion by the majority
people. If it's not stable, it's not a MIST item.
--
pjac...@ic.net
In article <333ECA...@clickable.com>,
Ansel Morgan (aka Christopher McKinstry) <in...@clickable.com> wrote:
>
> Philip Jackson wrote:
>
> > You have yet to prove or substantiate any of your claims. Put up or shut
> > up. A claim that all this will happen eventually, without any
> > substantive description of how, is just as vacuous and unsubstantiated
> > as all your other claims.
>
> about it. just bite your tounge, ignore my claims and put some data in
> the MISTIC database.
>
> it doesn't matter if my claims turn out to be true or not. the data is
> important.
Unfortunately the data you are collecting will largely be useless,
nonsense and pseudo-facts.
It is important in the interests of science to challenge your claims,
else what is to prevent people who make false claims from corrupting
science? I'm not worried about it and I won't ignore it. Rather I'll
devote a small fraction of my time to challenging your unfounded claims,
even if you ignore my constructive criticism.
Phil Jackson
-------------------==== Posted via Deja News ====-----------------------
http://www.dejanews.com/ Search, Read, Post to Usenet
pjac...@ic.net
In article <333ECD...@clickable.com>,
Ansel Morgan (aka Christopher McKinstry) <in...@clickable.com> whined:
>
> Jim, i will no longer respond to either you or phil jackson on this
> matter.
I don't care if you respond or not, since your responses to date have
been vacuous. In the interests of science I will continue to challenge
your claims, even though you ignore my constructive criticism.
> you two have spewed garbage on the comp.ai groups for years now,
> and have never produced a single thing.
You have amply demonstrated your incompetence to evaluate.
> neither of you are qualified to comment.
The quality of the comments that have been provided is sufficient
evidence of qualification to comment. Unfortunately you are not qualified
to respond, as evidenced by the feebleness of your responses.
> as for my ego being big, damn right.
> - how many tv interviews have you done? i've done well over 100.
> - when was the last time your were on CNN? january 8, 1997 for me.
> - how old were you when you made you first million? i was 17.
> - when was the last time you were asked to be keynote speaker at a
> convocation? february 28th, 1997 for me.
> - do you drive a jaguar? i do.
Assuming that these statements are true, they are irrelevant.
> remember that. i do, you (and phil) spew.
No. You spew and you are not doing anything for AI with MIST. I'm merely
challenging your claims, in the interests of science.
pjac...@ic.net
In article <01bc3d3d$40effba0$03d3c9d0@wjh_dell_133.dazsi.com>,
"Bill House" <bho...@dazsi.com> wrote:
>
> Philip Jackson <pjac...@mail.ic.net> wrote in article
> <333DA9...@mail.ic.net>...
> >
> > [snip]
> >
> data,
> > > > and effectively creates a scale of intelligence for any given test.
> >
> > Sorry, I have to disagree. It appears that the MIST collection of
> > true/false pseudo-facts will retain substantial subjectivity, that it
> > will not provide any valid quantifiable data, that it will not
> > effectively enable testing for intelligence.
> >
>
> automatically produce a human-level intelligence in a NN, I have to wonder
> how much the validity of the factoids really matters. In the case of
> humans, it's pretty obvious that people can believe the most outrageous
> fantasies to be factual, but as long as they manage to survive long enough
> to produce and raise progeny, they are able to persist. Untruth gets
> dressed up as culture and tradition and life somehow goes on.
The important thing is that despite having some incorrect beliefs, humans
are able to reason correctly about many things and solve many real world
problems sufficiently to survive and prosper. This indicates the value of
contextual reasoning in limiting the effect of incorrect / contradictory
beliefs -- again, something MIST does not address.
> If some kind
> of true AI was given the MISTIC database as an input, the real test would
> be how long it took for it to recognize the invalid factoids the database
> contains. Perhaps good BS-detection is a component of intelligence?
Yes -- Hence I suggested that if a True AI was given the MIST database,
it's response would be to add the statement "MIST is Stupid: True".
Cheers,
Bernard Tocail
Ansel Morgan wrote:
>
> Philip Jackson wrote:
>
> > You have yet to prove or substantiate any of your claims. Put up or shut
> > up. A claim that all this will happen eventually, without any
> > substantive description of how, is just as vacuous and unsubstantiated
> > as all your other claims.
>
> what i claim is tangental to the project of data collection. don't worry
> about it. just bite your tounge, ignore my claims and put some data in
> the MISTIC database.
>
> it doesn't matter if my claims turn out to be true or not. the data is
> important.
> K. Christopher McKinstry
> Homepage: Http://www.clickable.com/employees/chris
> World's Largest AI Project: Http://www.clickable.com/mist_corpus.html
> World's Coolest Interactive Drama http://www.cr6.com
test de réponse
Bernard Tocail
Christopher McKinstry wrote:
>
> Chris Hooley wrote:
> > The only other thing is that your mind will be affected by the medium
> > of collection. It will most resemble an adolescent boy. :-)
>
> there's a whole pile of items in the db like the following:
>
> do you feel lonely at time? true
> when peole laugh at you do you sometime feel bad? true
> the older you get the more mortal you feel. true
>
> and so on... thus, the human emotional factor is also captured in
> MISTIC. there's some very advanced stuff in there. it is becomming very
> human. a complete adualt consciousness can be captured with enough MIST
> items... not just an adolescent boy. (why you specified a sex, i have no
> idea).
> K. Christopher McKinstry
> Homepage: Http://www.clickable.com/employees/chris
> World's Largest AI Project: Http://www.clickable.com/mist_corpus.html
> World's Coolest Interactive Drama http://www.cr6.com
ffffff
pjac...@ic.net
In article <333FF4...@clickable.com>,
"K. Christopher McKinstry" <ch...@clickable.com> wrote:
>
> Philip Jackson wrote:
>
> > No, a coin cannot. Such statements are context dependent (a concept
> > McKinstry apparently doesn't understand, or else boneheadedly refuses to
> > address), and a person can answer them differently depending on the
> > context of a conversation. To answer such questions validly in a given
> > context requires human intelligence, and can't be done by coin flips
> > just as well as a person.
>
> phil, i'm not going to respond to you in public ever again.
I don't care whether you respond or not, since your responses have been
vacuous. It is clear that you simply cannot substantiate your claims nor
defend your ideas in a public forum.
> however, if you had read the MISTIC materials, which it is obvious you
> haven't, you would know that items such as "are you male?", are not
> answered in a consistent fashion by most people and thus are not
> permitted in a MIST test. Such questions were quite specifically
> excluded from the very beginning.
The point is I read your "materials" and have criticized the exclusion of
context-dependent questions from the beginning, and you have boneheadedly
ignored this criticism from the beginning.
>
> MIST is about the MINIMUM Intelligent Signal. A single binary test in
> complete isolation. MIST is the smallest test you can perform on an
> intelligent entity or an entity proporting to be intelligent. "Are you
> male?" is not, and cannot be a MIST item. on it's own, independent of
> any other questions, it tells you nothing about the consciousness of the
> entity being questioned. the entity could be a coin, and you wouldn't
> know the difference.
MIST is not an intelligence test at all. It is certainly a minimum test
-- You would be more accurate to call it the Minimum Index for Stupidity
Test.
>
> A MIST item must be stable, first and formost. That means that it must
> be answered, in isolation in a consistent fashion by the majority
> people. If it's not stable, it's not a MIST item.
Again, you just don't get it. It is said that "consistency is the
hobgloblin of small minds" and it certainly appears to be true of you.
Testing for things that a majority of people agree on is not a test of
intelligence at all, nor of commonsense since it lacks context
dependence. If 60% of the people contributing to your nonsense-base
believe Elvis is alive, does this mean it is "intelligent" to believe it?
The fact that *you* are the one judging whether most people would reply
the same way makes the whole thing complete nonsense.
Phil Jackson
--------------------------------------------------------------------
"...for the word is the sole sign and the only certain mark of the
presence of thought hidden and wrapt up in the body..." -- Descartes
--------------------------------------------------------------------
Standard Disclaimers. <pjac...@ic.net>
-------------------==== Posted via Deja News ====-----------------------
Leo Smith
> test de r=E9ponse
Ici un reponse
Comment voulez-vous un reponse dans cette newsgroup?
uk.test peut-etre?