In a recent forum someone stated that AI developers had not reached their AI goals due to the lack of a large database of knowledge. The response was that a large database of knowledge was not necessary to accomplish AI goals. This led me to observe that when one does not distinguish knowledge from intelligence then one cannot determine the value of knowledge to intelligence.

I believe our multiple uses of the term AI has led to its ambiguity. To help us clarify our meanings, I suggest we use the following additional terms: AHAI, AHAK, and AHAB, meaning Artificial Human Adult Intelligence, Artificial Human Adult Knowledge, and Artificial Human Adult Behavior.  This may offer a different perspective on testing, as well.

I leave it to others to find terms for projects with goals outside Human Adults.

I agree your words “one does not distinguish knowledge from intelligence then one cannot determine the value of knowledge to intelligence.” The main point is “the value of knowledge”. We can consider knowledge as material.  How to choose and organize knowledge to build intelligence is the key problem.

I believe we can build a finite seed knowledge base K. Based on K, AI can acquire the other necessary knowledge and intelligence independently or with limited supervision. The size of K is affected by the type of knowledge in it and how to organize the knowledge. We can build K with an affordable cost if we choose the right knowledge and find the right way to organize these knowledge.

The day we build K is the start day of AI age.

The following questions were copied from the Certification section of the Human Mind Map site at http://mindmap.iwarp.com/:

Test ID: 1 Coverage: Remember, recall, classify, interpret, translate.
Given: [Remember] Tom is a human.
Q: Describe Tom.
A: Tom is a human.

Test ID: 2 Coverage: Remember, recall, classify, interpret, translate.
Given: [Remember] Humans are mammals.
Q: Describe humans.
A: Humans are mammals.

Test ID: 3 Coverage: Remember, recall, classify, verify, interpret, translate.
Given: [Remember] Humans are mammals. [Remember] Tom is a human.
Q: Is Tom a mammal?
A: Yes, [Tom is a mammal].

Are these tests of AHAI, AHAK, or AHAB?

It’s my considered opinion that you’ve presented an example of AHAK, with a tiny amount of AHAI thrown in, insofar as you’ve related some information to the entity, and that entity acted on a correlation of that information with “one degree of separation”. Once someone can create an entity that can be given a Dell Logic Problem of the type you find in those word-search/crossword books that you can get at Walmart, then you’ve moved beyond AHAK, and fully into AHAI. And at that point, I’ll be cowering in the corner, waiting for the “Robot Apocalypse”.

Test ID: 1 Coverage: Remember, recall, classify, interpret, translate.
Given: [Rule] if &A memberof &B then translate to &A is a &B.
Given: [Rule] if &A is a &B then interpret to &A memberof set &B.
Given: [Remember] Tom is a human. (I will remember Tom memberof set human.)
Q: Describe Tom. (I recall Tom memberof set human.)
A: Tom is a human. (I am translating Tom memberof set human to Tom is a human.)

Now that we can “see” what the entity is “thinking”, is it AHAI, or AHAK?

@Robot spelled backwards man:

Test ID: 0001b Coverage: Remember, recall, classify, interpret, translate.
Given: [Remember] World of A.I.
Q: Describe World.
A: World of A.I.

Test ID: 0010b Coverage: Remember, recall, classify, interpret, translate.
Given: [Remember] World of A.I. is not enough.
Q: Describe World of A.I.
A: World of A.I. is not enough.

Test ID: 0011b Coverage: Remember, recall, classify, verify, interpret, translate.
Given: [Remember] Humans are mammals. [Remember] World of A.I. is not enough.
Q: Is World of A.I. not enough?
A: Yes. [World of A.I. is not enough.]

I’ve been trying to figure out how I could create an example to use for testing.

Given: Tom is a human.
Given: Tom is a male.
Given: Tom is a sibling of Gary.
Given: Tom is happy.
Q: Describe Tom.
A: ???

What is the answer?  If there is a large body of knowledge, how much is used to describe Tom?  Would “Tom is alive” be an appropriate answer?

While I can see that with macro (templates) you can simulate membership in a set, I don’t understand the reference to “thinking”.  Consider being in the blocks world and requesting (Q:) Put the small red pyramid on the big blue block.  (Given:) The table could contain the set of things; blocks, cubes, pyramids, boxes, etc. (A:) To “think”, the robot would have to find the specific pyramid on the table and the block and make sure the identified block was clear enough to have a pyramid placed on it. It would have to remove the pyramid from the table and add it to the things on top of the block. I’m not sure your macros like in the example you present could handle all these steps.  Yet this blocks task is more like “thinking” than reporting that a small red pyramid is a thing (on the table.)  If the things were Lego pieces and you got the robot to build a car and a garage, then we might be talking about AHCI (Artificial Human Childish Intelligence.)

Given: Tom is a human.
Given: Tom is a male.
Given: Tom is a sibling of Gary.
Given: Tom is happy.
Q: Describe Tom.
A: ???

What is the answer?  If there is a large body of knowledge, how much is used to describe Tom?  Would “Tom is alive” be an appropriate answer?

A: Tom is a happy human male sibling of Gary.

There is no other body of knowlegde given, only the information in the test is available.

author=“Gary Dubuque” date=“1311384946”]I’ve been trying to figure out how I could create an example to use for testing.

Given: Tom is a human.
Given: Tom is a male.
Given: Tom is a sibling of Gary.
Given: Tom is happy.
Q: Describe Tom.
A: ???

What is the answer?  If there is a large body of knowledge, how much is used to describe Tom?  Would “Tom is alive” be an appropriate answer?

While I can see that with macro (templates) you can simulate membership in a set, I don’t understand the reference to “thinking”.

You have correctly identified the point of the question.
Are remembering, recalling, interpreting, translating, and classifying  functions of “thinking”, i.e. intelligence.

I think that’s a part of intelligence, but not the whole, nor a majority of the whole. While it’s true that you cannot have intelligence without remembering, recalling, interpreting, translating, or classifying knowledge, if that’s all that you have, then it’s still not intelligence. I think that a certain amount of creativity (at the very least) is also required, as well as other, less easily defined qualities are required, too. At this point, I can’t say what these other qualities are, but I’ll think on it, and see if I can’t “put my finger on it”, so to speak.

It is true that these functions do not include all of the functions of intelligence.  Is the test useful in determining if the entity has the ability to learn from a tutor?

Toborman - Jul 23, 2011:

A: Tom is a happy human male sibling of Gary.

There is no other body of knowlegde given, only the information in the test is available.

You can’t be serious.  There has to be a program and all programs are information.  For example, you used happy and human and male as adjectives in your sample response. That came from the program having information.

Just as easily a male sibling could be reduced in that dictionary which allowed the words use as adjectives, reduced to the word “brother”. So the sentence with the same meaning could have been: “Tom is Gary’s happy brother.” Still no new information?

One of the tests on the site you referenced suggests that if a statement like
Given: “Tom is not a quadruped.”
is provided, the program when given
Q: “Is Tom a quadruped?”
will respond with
A: “No”.

But “No” is not information available in the test according to your stipulation.  Nor is the inference that if “a thing is not”, then “a thing is” produces “No” (not always true, i.e. happy and unhappy at the same time.)  Further testing examples for this CIA, “Part 2 - Learning categorical propositions by discovery (inference).” directly leads to being able to infer that when “Tom is happy”, he is also “alive”.  Inference can expand information quite rapidly.

I would avoid using this CIA stuff for determining intelligence.  It deals too much with isolated effects hinting at specific abilities and ignores the whole set of capabilities working together which is the core of intelligence.

Gary Dubuque - Jul 23, 2011:

Toborman - Jul 23, 2011:

A: Tom is a happy human male sibling of Gary.

There is no other body of knowlegde given, only the information in the test is available.

You are right, of course. I applied a consolidation rule that wasn’t shown. Without consolidation your test would look like this:

Given: Tom is a human.
Given: Tom is a male.
Given: Tom is a sibling of Gary.
Given: Tom is happy.
Q: Describe Tom.
A: Tom is a human.
A: Tom is a male.
A: Tom is a sibling of Gary.
A: Tom is happy.

Thanks for catching that.

Gary Dubuque - Jul 23, 2011:

You can’t be serious.  There has to be a program and all programs are information.  For example, you used happy and human and male as adjectives in your sample response. That came from the program having information. [...]

Thank you Gary, I was just about to say the same, but you put it better than I could.

It is an easy trap to fall into, incorporating external information without realizing it. We use so many pieces of information to complete even the simplest of extrapolations: memories/imagined scenarios/generalizations of behaviors and object properties. (Knowledge united with creativity, as Dave mentioned.)

As an aside, the word “creativity” is rather vague. It suffers from the same problem as the term “consciousness”—even it’s (many) definitions must incorporate still more vague terms in order to be satisfying. What would you say is a solid definition of creativity, a definition that only incorporates words and terms that have clear meaning and/or point to specific actions and behaviors?

EDIT: The following is more long-winded than I intended. Apologies.

I would say human creativity includes:
- Using incidental examples of object behavior to classify objects into broader categories.
- The ability to add/remove/partially add an object to a category, even based on information that is perhaps not related to the category directly, but is related to other information about the category’s items.
- The ability to use categorical information to determine/guess the outcome of a scenario involving categorized items which has never taken place (or at least been witnessed).
- The ability to change an object’s categorical information and, as above, determine/guess the outcome of a scenario. This ability could be used to update categorical information based on new information (scenarios) it experiences. Or, less practically, be used in artistic endeavors.

A simple example:

Given:
The cup is made of glass.
The cup feels smooth and cool.
The vase is made of glass.
The vase fell and then broke.
A fish tank can feel smooth and cool.

Guesses:
(1) Glass objects can feel smooth and cool.
(2) Glass objects can fall and then break.
(3) If a fish tank fell, it might break.

If a bot used this kind of categorical thinking, would you consider it creative? What if the “guess” output was much more complex, involving several objects and their categorized properties? At what point would (or could) the output be indistinguishable from human creativity?

Bonus: Back to the original reason for this post. Is there any implied but unstated information used to make the “guesses” in my example?

Toborman - Jul 20, 2011:

This led me to observe that when one does not distinguish knowledge from intelligence then one cannot determine the value of knowledge to intelligence.

How would you define the difference between knowledge and intelligence or between Artificial Human Adult Intelligence, Artificial Human Adult Knowledge, and Artificial Human Adult Behavior?

To me, the difference between knowledge and intelligence is the simple premise that knowledge is having data available for recall, but intelligence is the ability to use that data in different ways to solve problems or perform tasks or functions that simply having said data alone cannot accomplish. Anyone can learn that Ohm’s Law is V=I/R, but it takes intelligence to know just what that means, and be able to put that formula to use to build a circuit.

Of course, this is just my opinion, but…