November 1, 2006 10:47 AM

Artificial Intelligence is one of those phrases which defy definition. Just like Intelligence.

 

At one time, it was thought if a machine could play chess it was proof of AI. Then after  the algorithm was figured out, the definition was moved to if the machine could beat a grand master then it was proof of AI. Then solving real world problems, this was solved using lisp with a robot that could put the red cube on top of the blue cube, where the red cube was under a green cube.  Then nural networks, and fuzzy logic came along, which made it more general and able to solve some pattern matching problems (facial recognition for example) or controlling machinery within unclear parameters, and doing it precisly. The line keeps moving back and the more we learn about AI the more we realize just how complex the world is that we live in.

 I think that AI is wrong headed. Instead of defining it as compared to some human or animal ability (which are very very general) it should be defined as intelligence for a particular task.  So by "my" definition, AI is present for playing chess, facial recognition, and so on. It is not present for the hundreds (thousands?) of other tasks that we take for granted everyday.

One of the things that Mark Tildon has done is realized that many of the generalized behaviours that occur in nature, are a function of the body of the robot, and not a function of the cpu inside.  That is why his beam bots seem much smarter then they are.

I think that we are on the cusp of some very interesting times in robotics. It reminds me of when Radio Shack and Apple had brought out their first computers. They couldn't do much but they were the start of a new era. 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 1, 2006 12:36 PM

Well, I don't know much about Mark Tildon, but I belive that this was once his baby. http://www.beam-online.com
I first heard of him when I lived in Waterloo and he was at the Waterloo University. Apparently he would build little robots out of old printers and throw away stuff.  

I think the turing test is only good at showing how guilible people are, not how smart AI is. Or to put it another way, the turing test is a good example of a very narrow form of Intellegence. (back to my definition).

 As far as WowWee in a rating of 1 to 100 where 1 is dumb as a rock and 100 is average person, I think it's far below an insect.  Insects have better vision/taste/recognition systems.  Better communications systems and better mobility systems. They also are able to reproduce.  Of course insect's intellegence comes in large part from their bodies (vision/taste/mobility/reproduction etc). Without those they cannot communicate, recognize, defend etc.  

One thing though that WowWee's robot's have over the insects is that they can learn. Insects are very hard wired.

An experiment was done where a certain beetle would get food for it's eggs, go to the tree, put the food down, dig a hole for the eggs and deposit the eggs and food into the hole, then cover it. (I think there were more steps but I don't remember clearly). The experiment was after the food was put down it was moved a couple of inches. The beetle would loose it's place in the order of things and have to start from the beginning. The researcher did it several hundred times, sometimes moving the food so the beetle would have to walk over it to get back to where it thought it was. The beetle could have just picked up the food and carried on, but instead it started over from the beginning. No learning ever occured. Then again now that I think about it, maybe the beetle thought that since the food had been tampered with it was no longer suitable for it's eggs. (only the best dung more my eggs!).

It's hard to know motives for sure. But I remember that this experiment showed pretty clearly that this particular beetle could not learn any new paths in it's programming.

 Of course the learning by WowWee's robots is all the teacher to student kind, little of the learning on their own kind so again it's an interface issue. WowWee's robots have a better interface to be taught then insects do. 

 I really think that AI is as much a function of I/O as it is of CPU. Probably a formula like

AI = I * O + CPU 

 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 16, 2006 10:15 AM

Well, here's a thought. (Re reference points).

1. The robot takes a picture.

2. The robot moves slightly.

3. Takes another picture.

4. Xors the two pictures, you now have the differences between the two pictures as the third picture.

5.a    if memory is getting tight, then delete picture with the lowest reference count.

5.b  Convert to Black and white and reduce the amount of gray to 16 or fewer levels, and you can pack that picture pretty tightly into a storage area. (differences are small compared to a picture).

6. Xor against any previously stored pictures to calculate which has the least difference. The one that is mostly 0's is the closest.  You've found a reference point. Inc that picture's reference count.

7.  go to 3

So now you've got a reference list, you can do mapping software, (ie which pictures are close to which other pictures).  Your robot can traverse a house with ease.   

 I'd like to do something like this with the RS-Media, if the software and hardware specs were available. 

 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 16, 2006 12:15 PM

"At what point does the bot cross the line between pre-programmed reaction to information it's gathered and "intelligence"?"

 I think you are missing the point.   Intelligence is not a thing, it's a continum. (sp?)   The pre-programmed reaction is intelligence.  So my little algorithm can be used to increase the intelligence the robot exibits, but is not "intelligence" by itself.  

 

Using my algrithm, you could for instance.

1. Tell the robot that "This area is the kitchen".

2. The robot walks around scaning the kitchen.

3. When the robot enventually leaves the kitchen, you tell the robot that "this is not the kitchen."

4. When the robot goes to another room of importance you tell the robot that "this is the dinning room".

 and so on.

Then you could say to the robot, go to the kitchen, and it should be able to connect the pictures to define the path to go to the kitchen from the dining room, or any other rooms it has learned about.  It can even at this point figure out new paths since even though it traversed from the kitchen dining room, living room, bedroom, it might realize that the kitchen room pictures can be found close to the dining room pictures. 

 This is all due to caluclations (look up pathfinding in google) but will in theory exibit intellegence.

 

Of course, in theory, practice and theory are the same. In practice they are different.

 

 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 16, 2006 4:39 PM

I would disagree. If you use a remote control to navigate a car, that is obviously intellegence. (yours).

If you program a car, same thing.

Your last point, I would agree seems more intellegent on the part of the car. But consider this; when  you are doing the driving in either of the first 2 cases, the car still has to have the intellegence to follow your instructions.  

I understand that it's not thinking "Oh I guess I'd better do as he says", but it still is showing some intellegence.  In your last case, I would say it shows more intelligence because it involves more decission making. But it still is a difference of degree, not a difference of what. These kind of discussions can go around and around since (back to my first point way up there) there isn't a clear definition of what intelligence is. So I try to avoid these discussions by stating that all things are intellegent it's just a matter of degree. This sidesteps the problem of "what is intelligence?" 

So you've got a degree from "intelligence of a rock" to "intelligence of a dolphin" :)

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 17, 2006 1:44 PM

"Do you have anything other than anecdotale evidence for the experiment with the bettle?"  Pardon? I don't know what you are talking about.  If you are talking about my algorithm, no it's only an idea at this  point.

 The problem with saying that the ability to decide not follow uses the word "decide" which implies intellegence. You can't use a definition to define itself.  I state that all things are intellegent to a degree, including the car being pulled by a string (which would be a very low degree).  

 I agree that the one of the main problems with robotic intelligence is the lack of sensors. But on the otherside, the processing power to use the sensors to any degree is also significant. The processing power to drive motors is trivial. So most robots these days don't have the processing power needed.

 

As to your last question, "does it have to see and experience the world in the same way we do", I think up to the point that any of use see and experience the world the same it would, otherwise perceptions would skew interpretations which would also skew outputs which would further skew interpretations and so on. Think of the behaviour of different types of people who do have more or less the same experience, and you can see the effect is large.

 

 

 

 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 18, 2006 12:02 AM

Oh, ok I know what you are talking about now. I think I read about it in nature but I couldn't swear to it, and it was several years ago.   The point of the artical was that the beetle couldn't learn how to adapt to a slight change in it's routine even over many trials.  I probabaly shouldn't have used that example because I can't remember the particulars.

 Your previous post said  " I would say inteligence, starts with the ability to make a decision." which is what my comment was based on. As this thread deals with intellegence I thought your were offering a (partical) definition.

 I like your dirt idea. I normally use rocks, which have the same level of intellegence. I've seen rocks that have remained in place for years. :)

Perhaps instead of just disagreeing without really stating why (other then it doesn't make sense to you) you could offer your own definition of intellegence.  It's a very hard thing to do. I've offered a definition which although is unconventional, actually  encompasses every form of intellegence. I'll try to state it carefully. All things have intellegece, which is measured by their ability to do whatever they are good at.   Cheeta intellegence is in it's body design for speed. Human intellegence is in their ability to figure stuff out. 

I also gave an equation at one time. IQ = I * O + Brains

Notice that rocks don't have any IO, They may be the most brainy things in the universe but you would never know it because they can't demonstrate it.  Also they have on way to observe the world. So their intellegence may be limited to abstract thought.

This is a just thoughts that I've had over the years, and will probably lead to nothing, so to end on a more interesting note, here's a link to a robot that can actually do some thinking for itself. here

 

 

 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 19, 2006 6:19 PM

OK, (I'll play).  I/O should be a measurement of a difference between a normal output and what actually is output. (We can't measure input onlyt the rock can).  So a rock reflecting light, in an unusual way (LCD display for example) would be able to have a higher output.  A rock has a I/O that is as basic as an I/O can be. If we use a rock as the base level, then it has an IQ of 0. The beetle might have an IQ of 10, you spider 15, the bees 19, humans ?  As we still don't have a good measuring stick the numbers are all random. So given my equation, IQ = I * O + Brains, a rock has the IQ of whatever brains it has. 

Where it runs into more problems is that there are differnet types of IQ. A musician may have a high "music IQ" but low "math IQ".  This also shows up in learning.  Some kids learn better though touch, others through sight, others through sound. It all depends on the kid. So the problem is, two guys who are each brilliant in their own field, each of which is deemed beyond the normal person's ability. Which one has the higher IQ.  Well compared to each other, it's difficult due to the fact that they have many of the same attributies (can talk, walk etc), but have some special attributes as well.   I say they might be equally high, but in different areas. But then you've got an infinite number of areas, so therefore an infinite number of definitions. SO intellegence is as undefinable as division by 0, unless you specify the area. 

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 19, 2006 11:04 PM

I wouldn't measure it by your sleeping, but by the difference between playing music and sleeping. So going back to our rock, If you shine a light on it, and the light reflects back, that is an output. If the amount it reflects back changes, that is also an output.  The difference between the two is the output I'm looking for. Not the effect itself but rather the change in the effect.  As the amount of light reflected from rock doesn't change the output is zero.  What you are doing is saying that a voltage is an output, I'm saying a change in voltage is an output.  Listening to a speaker hooked up to a battery doesn't produce any sound (other then the initial change when hooking it up). Even though there is energy going through the speaker. Even though sound can reflect off the speaker. The speaker's output happens when the voltage going through it changes.

Therefore A rock has an IQ of brains (which is unknown) in my equation   of IQ = I * 0 + brains.  However I think brains is also a function of I/O and perhaps something else.

Just to be sure I'm complete, the subjective parts of the equation cannot be removed, since that is the way I perceive the world. The assumed parts of the equation cannot be removed since the equation itself is an assumption based upon my perception of the world.  How could you measure unassumed and nonsubjective elements when there are none? Consider that everything you know about the world is subjective and assumed. For example you assume that 2 plus 2 will equal 4 because you have always perceived it to work that way. There is no absolute proof, since you might be totally delusional and only be able to perceive it that way, although it actually equals, 100000000000004 (you just don't see the other numbers). 

:)  This is fun

On another note, since I have your attention. Have you made any progress in loading code onto the system rs-media. or doing any other fun stuff. (I've been busy and haven't really had a chance).

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?

November 20, 2006 9:27 AM

I took a course on AI for my masters and I'm giving you the same arguments I gave the rest of the class in the discussion. (Online learning is FUN). It was a great course, we learned about nerual networks,path finding, fuzzy logic, and all the while this thread was going on in the background. Some people thought I was insane, others came around :)

I've never thought of I/O as a constant, and that you would takes me by surprise. So when you are looking for your RS-232 are you looking for changes in voltage or the presense of a voltage? You are right that the brain is assumptive. I can't believe our thoughts are just a representation of the I/O we have had throughout our lives. But because inteligence must be perceived in order to exist, (at least to most people) and because inteligence cannot exist in a vaccum, (eg kittens raised in a box with no visual clues are virtually blind when they grow up) I strongly think that I/O is part of the equation.

If you need some code compiled for the arm, let me know I've got a working setup and can zip it off fairly quickly. (Getting it onto and off of the RS-Media is my problem).

All things considered, a byte is 8 times smarter then a bit. If that were true, then a byte should only store numbers up to 8. Could it be that a byte is greater then the sum of it's parts?