My robot can LEARN, but it can't UNDERSTAND

Am I making any sense ?

Yes I understand :)

The machine has no model of the world that it can use to generate such an understanding from.

The weakness with an ANN is it cannot explain why it categorises images as "blocked" or "clear".  It cannot even list the features it is using to determine this.  They are working on ways to remedy this.

Something I might try on my robot base is using sonar to avoid collisions.  I will think of an algorithm to achieve that task and program that into the computer. The computer will input the sonar distance data and processes it according to a set of rules given by me.

Another method might be to allow it to learn. First you need a signal to punish or reward its actions for a given input. Maybe a bumper switch would do the job. The machine would associate an input/output action with a collision.  I am guessing this will be a low distance sonar reading just before a collision.  It would then learn not to make certain moves when it gets a low distance reading.  Instead it will trial other moves perhaps based on a set of sonar reading in different directions eventually maybe learning to turn in the direction which has the highest distance reading?

Learning involves time and computing resources whereas to hard code a behaviour might use very little in the way of computing power. Maybe I will put my sonar hardware built earlier on the robot base and give it a go?

587

Note it doesn't end well for this little bot :)

https://www.youtube.com/watch?v=EJwNwbXVTCk

My robot can LEARN, but it can't UNDERSTAND

Am I making any sense ?

Makes perfect sense to me.   This is why I'm creating (for my own sanity) a distinction between A.I. (an ability to learn without understanding) versus R.I. (an ability to learn with understand).   Of course these are just definitions, whether we can build a machine-based R.I. is another question entirely. ? 

But yeah, I agree there's a huge difference between learning how to do something versus understanding what it's all about.

And this can be scary in terms of advanced A.I.  If we build robots that can do things but don't understand what they are doing that can indeed be quite dangerous.

So my goal is to try to build R.I. not A.I. ? 

Whether I could ever achieve this goal is whole other question.   But the goal can be well-defined, I think.

It basically comes down to whether the machine understands what it's doing or doesn't.  Pretty simple distinction I think.

Learning is a change in behavior over time for the better (as judged by some criteria).

As personal observation using empirical (however anecdotal) data 

The latest version of my ROS has applied a "fix" to the "learning mode" of my robot

I can now put the robot in a position on the floor with something in front of it, and take a picture of the thing being in front of it, and save the picture to a folder labeled "blocked"

Then I can remove the thing in front of it, giving it free space to move in a forward direction, and take a picture of the space in front of it being free to move in a forward direction, then save the picture to a folder labeled "free"

Then I can set the thing on a table near the edge, and take a picture, and put it in a folder labeled "blocked"

(Although, once I transfer this "brain" into my actual robot, I can't imagine a condition where it will be on a table, I am doing this anyway... Just in case. It might go outside and try to drive off the curb of the sidewalk, who knows. I'm trying to be thorough)

Once I've taken about 200 pictures for each category, I'm supposed to move that "dataset" (that's what they call it) to another folder that will do some crunching, which will apparently take a few hours according to the documentation), then I take the output from that crunchiness, and import it back into the robot, at which point, it will, theoretically, not run into things, or try to drive off a table that it should never be on in the first place

I'm "training" the robot. It's "learning" not to drive off a cliff or run into things (I'm going to let it run over small things tho)

The robot will "know" certain conditions where it's free to move, or blocked from movement...

This would be exhibiting a behavior that shows a positive improvement

The robot is situationally aware of its immediate surroundings, but only as far as "move" or "not move" is concerned

This does not mean, however, that the robot knows what will happen to it if it does decide to drive off a cliff

Yes, it has learned NOT TO drive off a cliff, but it doesn't have an understanding of WHY it shouldn't drive off a cliff other than its programming has told it not to

My robot can LEARN, but it can't UNDERSTAND

Am I making any sense ?

I don't have a fixed view about any approach but if I can't question a particular approach then no exchange of views or knowledge can take place.

It didn't sound to me like you were asking questions.   It sounded to me like you were attempting to make an argument that words are not the same thing as the nouns and verbs, etc, that they label.

If that was supposed to be a question then my answer is this:

Semantics is about the "meanings of words", not just about words themselves.   And this is the main idea behind my Semantic approach to A.I. that I have been talking about.  So it's not just based on words, it's based on the meanings of the words (i.e. having an understanding of what the words actually mean).   Creating a Semantic A.I. that can do this is the challenge.   And it's not easy.   If it was easy Semantic A.I. would most likely already be the front runner in A.I. today.   But just because it's not easy is no reason to reject the idea as being impossible.   After all, humans associate words with their meanings all the time.   If they couldn't do that then words themselves would be meaningless, and thus useless.   So we already know it's possible.   It's just a matter of figuring out how to do it using a computer, and/or other techniques such as ANNs or whatever other tools might be available.

Posted by: @casey

I now remember reading about semantic networks a long time ago in an AI book.  After you first wrote about them I actually spent some time reading about them in case I missed something.

No one has yet been able to create a successful  A.I. based on semantics (at least not that I am aware of).   Although it is quite possible that Hanson robotics is using a semantic approach with their robots.  I don't know the basis of their A.I. architecture.

The article you referenced is only one approach.   In fact, in that particular approach all they are doing is trying to make connections between words rather than attempting to build a system that understands what the words mean.  While semantic nets or trees can be useful, they cannot be the entire picture since they don't include the actual meanings of the words.   In fact, your argument that words are not the same as the concepts they represent would actually apply to that particular method.    I've seen that approach to building "Semantic A.I." as well, and I too saw the limitations of that method.   It's an attempt to get at the meanings of words by associating  words with each other.  This method actually has panned out for those who are in the business of data analysis.  It is possible to discover semantic connections in large databases using this method.   But in the end, a human is still required to make sense of those  connections.  So it's not really an intelligent system on its own.   It's just a way of organizing words into patterns that may have some conceptual commonality.

Here's an example of these kinds of connections from the link you posted:

Notice that in this tree all that is being done is an attempt to make connections between words that can be associated with an emu.   While this type of word association can indeed be useful, especially for humans who are looking for semantic connections, there is nothing in this system that actually associates these words with the concepts they they represent.

For example, the word "fly" is associated with an action a bird can take.  But this system doesn't say anything about what "flying" actually mean.   The same thing for the word "wings", wings are simply something that a bird has, but this system doesn't say anything about what wings are, or what they might be used for.

As I have noted above, these kinds of semantic nets can be useful for humans to be able to make connections between data in large databases.   This is because the humans already know what all these words mean.  But if you want to use semantics as a basis for A.I. in a machine, you'll need to have more than just a these  relationships between words.    You need to also include the actual semantics (i.e. meanings) of these words.

Associating the word fly with the word bird doesn't tell us anything if we don't already know what a bird is and what flying is.

So there's far more to it than just creating trees of related words. 

Having said this, it should also be obvious that in any semantic system trees of related words will naturally arise, and they will indeed be very useful.   For example, the word "fly" and the concept it represents will naturally be associated with the word "wings" and the concept it represents.  In fact, in this example it's worthy to note that it wouldn't even make any sense to try to speak about "wings" in a meaningful way until the concept of "flying" has already been learned.   It would be extremely difficult to try to explain to someone what wings are if they don't already have an understanding of what it means to fly.

This is why I am very interested in the link I've provided about how humans learn semantics.

It supposedly takes a human about 2 years to get to the following point: Knows 50 to 100 words. Uses short, two- or three-word sentences and personal pronouns ("I fall down!" "Me go school?").

We are talking about a very small system here.  Only 50 to 100 words.   Yet we are already capable of building 3-word phrases that convey meaningful ideas.   These aren't just related words, but instead these are words that the intelligent mind understands in terms of what they mean conceptually.

In fact, I question the accuracy of  this information.   I can't imagine a 2-year-old understanding a concept of "Me go school"  and having access to a vocabulary of only 50 to 100 words.   My guess is that the kid is far more likely to have some understanding of more like 500 to 1000 words, if not more at that point.  I'm not sure where they came up with the 50 to 100 words.   Obviously they would need to guess at that based on various assumptions and observations.

In any case, even looking at a system that only has 500 to 1000 words defined and can create meaningful 3-word sentences appears to me to be a system that should be somewhat manageable.   Starting at the beginning and working forward toward that goal seems to me to be a vary promising approach. 

I would think that if you were successful in creating even a small system of a mere 1000 words where meaningful 3-word sentences were being formed by an A.I. that demonstrated that the A.I. "knows" what it's talking about based on the meanings of these words that would be quite the achievement.

And if you got that far, you should be in a position to have a good idea of how the system works.  Knowing this you should be able to create a system that can grow on its own using those principles.   So at that point the robot should indeed be able to "Me go school" and learn even more words and concepts.

If I can get a robot to a point where I can actually start teaching it like we teach a human child that would be quite awesome I think.

In any case, I'm not interested in arguing about this.   If you have no interest in a semantic approach to A.I. and machine learning that's perfectly ok.  No one is trying to convince you otherwise.   All I'm doing in these threads is sharing my approach.

I have had an interest in A.I. and an interest in how it might be implemented for a long time. I now remember reading about semantic networks a long time ago in an AI book.  After you first wrote about them I actually spent some time reading about them in case I missed something.

However in these posts I was writing about the theory of machine learning as I understood it not about semantic networks in particular.  I don't have a fixed view about any approach but if I can't question a particular approach then no exchange of views or knowledge can take place.

https://users.cs.cf.ac.uk/Dave.Marshall/AI2/node59.html

Anyway I will drop the subject and get back to coding my simple brained robot base ...

Language rides on concepts we already have derived from learning about our world via our sensors.

And I've already explained that when I'm speaking about Semantic A.I. I'm not limiting this to the words themselves but rather the concepts the words are being used to label.  Perhaps you didn't catch that part?

In any case, I'm not interested in arguing about this.   If you have no interest in a semantic approach to A.I. and machine learning that's perfectly ok.  No one is trying to convince you otherwise.   All I'm doing in these threads is sharing my approach.

If you have a different approach to A.I. and machine learning you are more than welcome to share the approach that you prefer.   This is what makes the world an interesting place. Everyone has their own ideas on how they would prefer to approach and tackle various problems.

So please feel free to describe your approach to A.I. and machine learning    That's the whole point of these forums. ? 

I can assure you there is no hand waving philosophy involved. 

Language rides on concepts we already have derived from learning about our world via our sensors. A child first interacts physically with its world to learn about it before it can learn to use a language.  Objects are probably experienced as features that move together in 3d space. Different objects have different features. Some objects have features in common thus belong together as a class. You cannot tell a child "this is called a book" if it doesn't already know what object you are referring to.  All you are doing at the beginning of teaching language is associating a word with an object.  The child has to make that connection perhaps by the word and object continually being associated as occurring at the same time. That is what I mean by language being grounded in the real world. Once the concept of a word being used to represent an object is made then the road to learning and using language begins.

Take the famous case of Helen Keller born deaf and blind.
https://en.wikipedia.org/wiki/Helen_Keller

Sullivan arrived at Keller's house on March 5, 1887, a day Keller would forever remember as my soul's birthday. Sullivan immediately began to teach Helen to communicate by spelling words into her hand, beginning with "d-o-l-l" for the doll that she had brought Keller as a present. Keller was frustrated, at first, because she did not understand that every object had a word uniquely identifying it. In fact, when Sullivan was trying to teach Keller the word for "mug", Keller became so frustrated she broke the mug. But soon she began imitating Sullivan's hand gestures. “I did not know that I was spelling a word or even that words existed,” Keller remembered. “I was simply making my fingers go in monkey-like imitation.”

Keller's breakthrough in communication came the next month, when she realized that the motions her teacher was making on the palm of her hand, while running cool water over her other hand, symbolized the idea of "water".

When we say someone or something is "intelligent" we are referencing their or its ability to acquire and apply information or knowledge which is a behavior.  It is the behavior that decides if the thing is intelligent not the thing itself.

You seem to be taking these concepts down an unnecessary philosophical rabbit hole.

I've already explained why behavior itself does not represent intelligence.   I can program a robot to behave in an apparent intelligence fashion.   And therefore the robot will exhibit what appears to be intelligent behavior.  But this doesn't mean that the robot itself has any ability to figure anything out or understand anything beyond what I had programmed it to be able to recognize.   Therefore I think it would be absurd to suggest that the robot itself is intelligent even though it clearly exhibits intelligent behavior.

So I simply reject your definitions.  They are not definitions that I find to be either meaningful or useful.

Posted by: @casey

My impression of a semantic net is that it involves a man made data base and is not grounded in the real world? If I ask the robot, "Is there a book on the table" it has to be grounded in the real world with a vision system capable of recognizing both the table and the book and their spatial relationship.

Again I simply disagree with your view on this.  I agree that words that humans have chosen to invent for the sake of communicating ideas and concepts are man made.  In fact, mankind has invented many different kinds of languages.   However, what I disagree with is  your suggestion that these semantics systems are not grounded in the real world.    To the contrary, I see them as being firmly grounded in the real word.   After all even language of different cultures can be translated in a meaningful way into other languages.   Why?  Because the whole idea behind these languages is to convey meaningful concepts and ideas that are indeed grounded in the real world.   So I don't understand your views on semantics here when you suggest that semantics isn't grounded in the real world.

Also, you go on to use an example of a robot being able to recognize a "book" on a "table".   But you are already using semantics to refer to these objects.   Does the robot know what a "book" and a "table" is?  Or has the person who programmed the robot just given the robot these terms to refer to various objects?

Could you fool the robot into thinking that a mere block of wood sitting on the table is a "book".   And would the robot  have any way of checking to see if that's truly the case?

Psychologists actually do this sort of thing with small children to test at what age a the human mind begins to  be able to recognize when things aren't as they had originally appeared to be.

In fact, I've actually been studying research that these scientists have been doing on young children with the hopes of using these same types of experiments and criteria to measure how well my robot is evolving in terms of its cognitive skills.

This is where the idle philosophy ends, and the real world begins.  We can actually design tests to see what a robot can actually understand and what it cannot understand. 

Here's a link to an article on the timeline of human child development in terms of learning to talk:

Your child's talking timeline

I found this timeline quite interesting and I'm going to see how well my robot compares with this.

Toddler 12 months:

Notice that it takes a full year for a human toddler to learn to use only one or two words in a meaningful way.   I'm hoping that my robot will get off the ground semantically much quicker than this. ? 

But it is important the robot knows what the words mean and isn't just saying them because they have been programmed into a speech engine.   And this is where the Semantic A.I.  model begins.  The robot has to know what these words mean.  Not just speak them because it was programmed to do so.

At 16 months:

Talks to someone much of the time as opposed to just babbling to no one in particular. Calls you to get your attention ("Mommy!"), nods and shakes head for yes and no. Makes many common consonant sounds, like t, d, n, w, and h.

At 18 months:

Has a vocabulary of about 10 to 20 words, including names ("Mama"), verbs ("eat"), and adjectives ("cold"). Uses common phrases ("want doll") to make requests.

Notice that at 18 months we're only working with a vocabulary of 10 to 20 words!  The keys issue is to fully know and understand what those 10 or 20 words actually mean.

Almost 2 years in:

Starts putting two-word phrases together for more novel purposes ("Daddy go," "milk mess").

Two years and we're only putting together very crude two-word phrases.

A solid 2 years in:

Knows 50 to 100 words. Uses short, two- or three-word sentences and personal pronouns ("I fall down!" "Me go school?").

Things are starting to accelerate.   If I had a robot that could construct meaningful 3-word sentences and actually know what it's talking about I'd be breaking out the champagne! ? 

These wouldn't just be 3-word sentences I programmed into a speech engine.  These would be meaningful sentences the robot is constructing to convey meaningful ideas.   Yep, I'd have champagne all over the place     at this point.

In about 3 years:

Can carry on a simple conversation about something in the immediate environment. Asks simple questions frequently. Expands phrases from three- to six-word sentences and develops a vocabulary of 200 to 300 words, including lots of verbs.

Uses past tense by adding a "d" sound to verbs ("runned") and plurals by adding an "s" sound to nouns ("mans"). Uses pronouns (I, she, we) correctly. <- shows signs of understanding the meanings of these words.

If I make it this far I will proclaim to have created a valid A.I. ? 

A robot that I can actually talk with and it will give me meaningful replies and understand the context of the conversation.

If I make it this far I'll consider the project to be a total success.

Not only this, but if I make it this far why should I think that this couldn't just continue to evolve?

3- 4 years:

Favorite words often include "why," "what," and "who." Can be understood most of the time. Can tell you what happened if you were out of the room.

4 - 5 years:

Communicates easily and can retell a simple story with a beginning, middle, and end while looking at pictures. Can use four to five sentences to describe a picture, with most of the grammar elements in place. Uses more than one action word in a sentence.

Pronounces most sounds correctly but may still have trouble with th, r, s, l, v, ch, sh, and z. Uses lots of descriptive words, including time-related words like "yesterday."

6 - 7 years:

Can describe how two items are the same or different, retell a story or event without the help of pictures, and recount past conversations and events. Uses some irregular plural nouns ("men," "teeth").

8 years:

Has mastered all speech sounds as well as the rate, pitch, and volume of speech. Uses complex and compound sentences correctly and is capable of carrying on a conversation with an adult.

Time to send the robot off to college! ? 

The word 'intelligence' is an abstract noun. It is often conflated with 'sentience' or 'consciousness' or 'soul' which people do imagine as being a thing in the brain.  It may feel that way but that is not evidence it is the case.

My impression of a semantic net is that it involves a man made data base and is not grounded in the real world? If I ask the robot, "Is there a book on the table" it has to be grounded in the real world with a vision system capable of recognizing both the table and the book and their spatial relationship.

I would say the motor skills are primary not secondary. Reasoning is the same as goal directed motor behavior and is supported by the same neural machinery.

Behavior is all there is apart from the thing that is behaving.

Agreed.  I take this to be an observation that the thing that is behaving must then be taken into account as well since it's clearly part of the equation.

Posted by: @casey

The higher levels have trained the motor modules to carry out the task all by themselves. 

Agreed.  And this is why I'm interested in studying these "higher levels" rather than focusing on the motor skills that are clearly secondary.