TimB - 09 March 2012 10:57 AM

We can “understand” things because we have complex verbal behavior.

We understand things because our nervous system was evolved for and we grew up in a three dimensional space-time continuum.  All of this verbal drivel is simply symboloism for that.  Talking like the verbal stuff is more important than reality is nonsense.  Reality is incapable of giving a damn about our symbols.  We mostly use them to confuse each other. 

The map is not the territory.

http://en.wikipedia.org/wiki/Alfred_Korzybski

The Tyranny of Words (1938) by Stuart Chase
http://www.youtube.com/watch?v=M9H1StY1nU8

psik

psikeyhackr - 09 March 2012 11:54 AM

TimB - 09 March 2012 10:57 AM

We can “understand” things because we have complex verbal behavior.

We understand things because our nervous system was evolved for and we grew up in a three dimensional space-time continuum.  All of this verbal drivel is simply symboloism for that.  Talking like the verbal stuff is more important than reality is nonsense.  Reality is incapable of giving a damn about our symbols.  We mostly use them to confuse each other. 

The map is not the territory.

http://en.wikipedia.org/wiki/Alfred_Korzybski

The Tyranny of Words (1938) by Stuart Chase
http://www.youtube.com/watch?v=M9H1StY1nU8

psik

You understand things because you have developed all of the basic verbal behavior abilities that a human needs.  Behavior is real.  Verbal behavior is a special kind of behavior.  It really exists also. (It can be measured/recorded.)  You don’t have a measurable operational definition for the concept of “understanding”.  But to the extent that you agree that “understanding” is something “real”,  and you have interest in a machine being able to have understanding, it would behoove you to have a measurable operational definition of “understanding”.  I merely suggest that is possible through utilizing the various components of verbal behavior.

(Skinner’s concepts that he developed in his book “Verbal Behavior” are currently being used effectively in helping some young autistic children develop complex verbal behavior.) It seems to me that examining that conceptual framework may be useful in the eventual development of a machine that can learn and do complex verbal behavior.

http://www.ijpsy.com/volumen8/num3/205/skinners-verbal-behavior-EN.pdf

psikeyhackr - 09 March 2012 10:33 AM

domokato - 08 March 2012 07:17 PM

Okay, if you want to talk about symbols or classifying things, look at Watson. Wouldn’t you say he has a rudimentary form of “understanding symbols”? He certainly understands wordplay and puns. That’s symbolic in the sense that he’s hearing one thing and taking it to mean another.

This is interesting.  Yesterday I watched the Prophets of Science Fiction: Arthur C. Clarke.  Because of HAL 9000 they had a segment about Watson and an interview of the designer.  He said that what Watson does is compare symbols to symbols and can use various algorithms to recognise when one set of symbols have a high probability of meaning the same thing as another set of symbols.

I believe our brains do something similar, unconsciously and very quickly, in most cases.

He admitted that Watson did not really understand what the symbols meant.  So it could essentially recognise when different people were saying the same things in different ways but have no idea what they were actually talking about.

That is why it could win at Jeopardy.  High speed searches and comparisons of symbols but not actually understanding anything.   

TimB’s response was good. I will just add that that’s why I used the qualifier “rudimentary” in “rudimentary understanding”. In my mind, it is definitely on the path to human-style understanding, and is a component of it. I think if you want to assert that there is something called “actual” understanding that is distinguishable from a “simulation” of understanding, you need to define your terms and justify why you draw the line where you do.

This should be relevant:

Scientists Claim Brain Memory Code Cracked

http://www.sciencedaily.com/releases/2012/03/120309103701.htm

psik

whoa, cool. thanks

TimB - 09 March 2012 12:15 PM

You understand things because you have developed all of the basic verbal behavior abilities that a human needs.  Behavior is real.  Verbal behavior is a special kind of behavior.  It really exists also. (It can be measured/recorded.)  You don’t have a measurable operational definition for the concept of “understanding”.  But to the extent that you agree that “understanding” is something “real”,  and you have interest in a machine being able to have understanding, it would behoove you to have a measurable operational definition of “understanding”.  I merely suggest that is possible through utilizing the various components of verbal behavior.

So you think a 3 year old that toddles around the house and opens and closes doors does not have an “understanding” of what doors are even if it has not learned the word “door”.

von Neumann machines have memory locations in their CPUs called registers.  A modern CPU might have 32 registers with 32 bits each.So that is only 128 bytes in the CPU.  The contents of these registers can be manipulated and moved according to machine instructions in RAM.  Now RAM these days can be in the gigabytes.But the CPU only executes one instruction at a time and there can only be 128 bytes of data in the CPU at a time.  But what can that data mean?

Each pixel in a camera these days is three bytes per pixel.  That makes 16,000,000 colors possible.  So an 8 megapixel camera uses 24 megabytes.  But only 128 bytes can fit into the CPU.  So how does a program sitting in memory control a CPU that can only hold 128 bytes analyse a picture that is 24 megabytes.  There is all kinds of finding edges and analysing shapes and things would get even more complex with stereo cameras trying to deal with three dimensions.  Not even 43 pixels can fit into the CPU at one time.

Plenty of people say the CPU is the brain of the computer but it is the program in RAM that is any supposed intelligence.  But all data is just a combination of bits.  A given combination of bits can be a certain color but that exact same combination can be a temperature or a wind speed.  It is all a matter of how the program treats the bits.

Here is the Kinect training:

http://www.youtube.com/watch?v=fQ59dXOo63o

The software memorised the responses to give to the shapes.  He could have called them anything.

psik

TimB’s response was good. I will just add that that’s why I used the qualifier “rudimentary” in “rudimentary understanding”.

mediafire movie

psikeyhackr - 10 March 2012 05:54 PM

Each pixel in a camera these days is three bytes per pixel.  That makes 16,000,000 colors possible.  So an 8 megapixel camera uses 24 megabytes.  But only 128 bytes can fit into the CPU.  So how does a program sitting in memory control a CPU that can only hold 128 bytes analyse a picture that is 24 megabytes.  There is all kinds of finding edges and analysing shapes and things would get even more complex with stereo cameras trying to deal with three dimensions.  Not even 43 pixels can fit into the CPU at one time.

Working data is stored in RAM. The values are copied into registers in the CPU so they can perform simple calculations on them, then store the results back into RAM (or disk for long term storage).

GPUs!!!

domokato - 11 March 2012 02:20 PM

psikeyhackr - 10 March 2012 05:54 PM

Each pixel in a camera these days is three bytes per pixel.  That makes 16,000,000 colors possible.  So an 8 megapixel camera uses 24 megabytes.  But only 128 bytes can fit into the CPU.  So how does a program sitting in memory control a CPU that can only hold 128 bytes analyse a picture that is 24 megabytes.  There is all kinds of finding edges and analysing shapes and things would get even more complex with stereo cameras trying to deal with three dimensions.  Not even 43 pixels can fit into the CPU at one time.

Working data is stored in RAM. The values are copied into registers in the CPU so they can perform simple calculations on them, then store the results back into RAM (or disk for long term storage).

I already said:

Now RAM these days can be in the gigabytes.

But information cannot be “seen” there.  There are no words in the language to explain it.  The only “awareness” is in the CPU and that memory is very limited.  von Neumann machines do not parallel process like the brain.  So what “understands” this so called information?  How can a program “understand” the data that cannot possibly fit into the CPU when the only place it can be analysed is in the CPU?

psik

psikeyhackr - 12 March 2012 12:21 AM

But information cannot be “seen” there.  There are no words in the language to explain it.  The only “awareness” is in the CPU and that memory is very limited.  von Neumann machines do not parallel process like the brain.  So what “understands” this so called information?  How can a program “understand” the data that cannot possibly fit into the CPU when the only place it can be analysed is in the CPU?

psik

psikeyhackr - 12 March 2012 12:21 AM

domokato - 11 March 2012 02:20 PM

psikeyhackr - 10 March 2012 05:54 PM

Each pixel in a camera these days is three bytes per pixel.  That makes 16,000,000 colors possible.  So an 8 megapixel camera uses 24 megabytes.  But only 128 bytes can fit into the CPU.  So how does a program sitting in memory control a CPU that can only hold 128 bytes analyse a picture that is 24 megabytes.  There is all kinds of finding edges and analysing shapes and things would get even more complex with stereo cameras trying to deal with three dimensions.  Not even 43 pixels can fit into the CPU at one time.

Working data is stored in RAM. The values are copied into registers in the CPU so they can perform simple calculations on them, then store the results back into RAM (or disk for long term storage).

I already said:

Now RAM these days can be in the gigabytes.

But information cannot be “seen” there.  There are no words in the language to explain it.  The only “awareness” is in the CPU and that memory is very limited.  von Neumann machines do not parallel process like the brain.  So what “understands” this so called information?  How can a program “understand” the data that cannot possibly fit into the CPU when the only place it can be analysed is in the CPU?

psik

Are you saying parallel processing is necessary for understanding? I don’t see why that should be the case. It’s not like the entirety of your visual data can fit into a single neuron either. Many neurons working together to process a bunch of data in parallel is functionally analogous to a single CPU processing the same data in serial.

Because the 1,500 cubic centimeter human brain is about 100,000 times as large as the retina, by simple calculation, we can estimate the processing power of a average brain to be about 100 million MIPS (Million computer Instructions Per Second ). In case you’re wondering how much speed that is, let us give you an idea.

1999’s fastest PC processor chip on the market was a 700 MHz pentium that did 4200 MIPS. By simple calculation, we can see that we would need at least 24,000 of these processors in a system to match up to the total speed of the brain !!

http://library.thinkquest.org/C001501/the_saga/compare.htm

psik

Peter Perepel - 11 March 2012 07:11 AM

TimB’s response was good. I will just add that that’s why I used the qualifier “rudimentary” in “rudimentary understanding”.

Language!  According to Dr. Hayakawa, language is the source of our superiority.  I was so fortunate to have taken a night course from Hayakawa when I was in High School.  We had access to the special lectures at UCLA and it was just a matter of getting a ride to and from the facility.  The timing of this was just after the end of WW2 and yet Hayakawa was one special professor.  I still have my text book.  How we use our spelling techniques, our punctuations as well as well defined words can make or break any communications. 

I will never understand the debate involving the Singularity but I love to read the posts from CFI on any subject.  This is a very special site.

Sandy Price - 13 March 2012 08:06 AM

I will never understand the debate involving the Singularity but I love to read the posts from CFI on any subject.  This is a very special site.

I feel exactly the same way Sandy. There is a lot of stuff here that is outside of my understanding but I usually still manage to learn something from every thread.