Great danes and chihuahas

"Robert"s post in another context reminded me of this and it always puzzles me. How does a Great Dane recognize a Chihuahua as another dog, even across the road?How, when we teach our toddlers, "doggie" in the picture book, how do they recognize one in the street.What is "doggieness" - clearly not just 4 legs, ears, eyes and tail. There is a kindof template "dog" shape which is wonderfully flexibile in our minds and the minds of dogs (and timourous cats).

You mean your kid(s) never pointed to a dog or cat and called it by the opposite name?While a "doggie" profile isnt hardwired into our hindbrains, facial recognition is.
As is the infant "scanning eyes" movement.
Its actually involuntary until the CC takes over./How, when we teach our toddlers, "doggie" in the picture book, how do they recognize one in the street.//
Ah the majesty that is the Cerebral Cortex!
If I told you that a supercomputer could do it in less time, would you beleive me?// "Robert"s post in another context reminded me of this and it always puzzles me. How does a Great Dane recognize a Chihuahua as another dog, even across the road?//
Canines all have a wonderfully powerful sense of smell.
Why would that Bowser *not* recognize little Chi Chi as another dog?On a side note... in the book Darwins Ghost by Steve Jones, he points out that in general the domestication of canines by humanity has resulted in dogs being bred to look like puppies and not like their wild bretheren.Its a favorable reaction to this "doe-eyed" look, akin to newborn humans, that is hardwired into all of humanity.Someone postulated that this was why the Area 51 aliens, the "greys" look the way they do... large eyes, receeding chin, child-like size...(hows that for running the topic gamut?)brett------------------
Faith in a delusional belief does not make that belief not delusional.

[QUOTE][b]
Someone postulated that this was why the Area 51 aliens, the "greys" look the way they do... large eyes, receeding chin, child-like size... [/QUOTE][/b]People inclined to buy aluminum foil in bulk will tell you they are designed to look like infant mammals to be disarming, but the much more interesting (and likely) explanation is that something in our psychology is causing the imaginary "Grey" to symbolize some unspeakable mortal fear or instinct associated with being or being around infants that lurks deep in the human subconscious. Why infants?

//Why infants?//
Two words.Toxic poopy.brett------------------
Faith in a delusional belief does not make that belief not delusional.

quote:

---

Originally posted by bretheweb:
You mean your kid(s) never pointed to a dog or cat and called it by the opposite name?

---

No. It was this that got me interested.
[b] [QUOTE] /How, when we teach our toddlers, "doggie" in the picture book, how do they recognize one in the street.//
Ah the majesty that is the Cerebral Cortex!
If I told you that a supercomputer could do it in less time, would you beleive me?[/b][/QUOTE]No I wouldn't believe you :-)This is very close to my line of work. Visual processing is the great failure of AI.
[b] [QUOTE]Canines all have a wonderfully powerful sense of smell.
Why would that Bowser *not* recognize little Chi Chi as another dog?
[/b][/QUOTE]Experimental observation by myself and my dog partly rule this one out, but not conclusively. I did think it was likely - they really do have the most astonishing sense of smell.
[b] [QUOTE]On a side note... in the book Darwins Ghost by Steve Jones, he points out that in general the domestication of canines by humanity has resulted in dogs being bred to look like puppies and not like their wild bretheren.[/b][/QUOTE]Indeed - domesticated dogs are just stuck in adolescence.

//No. It was this that got me interested.//
Ah, well theres the problem... your kids are too smart!
lolHonestly, I've heard this many times from cousins and sibs and whatnot.There is actually a developmental stage that they go through, linguistically, where categories arent terribly well defined, so indeed any 4 legged creature becomes "doggie".
They grow out of it as finer disctinctions are made.//No I wouldn't believe you :-)This is very close to my line of work. Visual processing is the great failure of AI.//
Good.
Now how about if I told you that I had a supercomputer that was the most complex and fastest in the known universe?How about then?//Experimental observation by myself and my dog partly rule this one out, but not conclusively.//
Hmmm... should I even ask?
lolI have a hard time believing that one dog could confuse another dog for something else.brett------------------
Faith in a delusional belief does not make that belief not delusional.

quote:

---

Originally posted by bretheweb:
There is actually a developmental stage that they go through, linguistically, where categories arent terribly well defined, so indeed any 4 legged creature becomes "doggie".

---

We're bilingual, so there may be a difference there.

quote:

---

Now how about if I told you that I had a supercomputer that was the most complex and fastest in the known universe?

​

How about then?

---

Em ... can I have job?

quote:

---

I have a hard time believing that one dog could confuse another dog for something else.

---

Except a trouser leg ...

//Em ... can I have job?//
Ha.
Too late.
You already created one.//Except a trouser leg ...//
Oh, hell, thats not confusion thats *expediency*. brett------------------
Faith in a delusional belief does not make that belief not delusional.

quote:

---

Originally posted by Mister Pamboli:
...No I wouldn't believe you :-)This is very close to my line of work. Visual processing is the great failure of AI....

---

Thats interesting Mr P, I have a few questions if you don`t mind...Q/Is visual processing an algorithmic process?Q/If it is does this imply that our "software" is more efficient than the software currently used by the supercomputer?Q/Or is the problem that the hardware of the supercomputer is inferior to human "hardware"?Q/Or do you think that it is because supercomputers lack a soul (though why this would affect visual processing and not ability to find large prime numbers I don`t know)?Q/Or do you think its something else?

Q/Is visual processing an algorithmic process?
It has been attempted algorithmically, but that approach has failed. There have been some interesting attempts at suing genetic algorithms which "teach" processors which features are more likely to be significant in a scene. But there are a number of models of how visual processing works in the brain - one of the most interesting ideas is that virtually everything we think we "see" is actually a construct of our brain which picks up remarkably little actual data and fills in the rest from experience and inference. This is being attempted in artificial visual processors by a number of people.Q/If it is does this imply that our "software" is more efficient than the software currently used by the supercomputer?
Q/Or is the problem that the hardware of the supercomputer is inferior to human "hardware"?
These are addressed I think in the previous answer. The hardware question is interesting however. It is the extraordinary flexibility of the human processing hardware that is so advantageous. The ability to use bits of the brain for something they normally don't process - you can almost "feel" this happen when you try to visually concentrate on something you haven't seen before while distracted by other more familiar visual stimuli.Q/Or do you think that it is because supercomputers lack a soul (though why this would affect visual processing and not ability to find large prime numbers I don`t know)?
How do you know I'm not a Turing machine answering this? No I don't believe computers have a soul - and I don't think it makes any difference to their processing ability.Q/Or do you think its something else?
When in doubt, go with this one!

To paraphrase Lewis Carol "interestinger and interestiner said joz" [QUOTE]Originally posted by Mister Pamboli:
[b]1)It has been attempted algorithmically, but that approach has failed. There have been some interesting attempts at suing genetic algorithms which "teach" processors which features are more likely to be significant in a scene. But there are a number of models of how visual processing works in the brain - one of the most interesting ideas is that virtually everything we think we "see" is actually a construct of our brain which picks up remarkably little actual data and fills in the rest from experience and inference. This is being attempted in artificial visual processors by a number of people.

​

2)The hardware question is interesting however. It is the extraordinary flexibility of the human processing hardware that is so advantageous. The ability to use bits of the brain for something they normally don't process - you can almost "feel" this happen when you try to visually concentrate on something you haven't seen before while distracted by other more familiar visual stimuli.

​

3)How do you know I'm not a Turing machine answering this? No I don't believe computers have a soul - and I don't think it makes any difference to their processing ability.

​

4)When in doubt, go with this one![/QUOTE]1)Thats interesting I assumed that it was an algorithmic method, say filtering by size, shape, colour, to reduce possible answers down to a subset (in the case of dogs the subset would be four legged animals of the appropriate size range) then searching this subset for a close match, If a match to some certain level of correlation is not found then the object would be labeled unknown and you search the whole set of known objects for similarities to add similar to information to the new entry....At least thats the way I thought it happened...2)So you think its a flexibility issue not a maximum capacity problem. I thought if hardware came up it would be something like this...3)Personaly as I subscribe to a functionalist view of the mind-body area of philosophy I only raised this as another possible explanation...As far as I`m concerned Turing machines already exsist they are human beings, its an artificial Turing machine that will be exciting....4)Wise words.....

quote:

---

Originally posted by Mister Pamboli:
There have been some interesting attempts at suing genetic algorithms which "teach" processors which features are more likely to be significant in a scene.

---

Q/Is binocular vision (and the associated depth perception) any sort of factor in this ability to assign significance? (or are we just talking 2D images here?)

quote:

---

Originally posted by joz:
Q/Is binocular vision (and the associated depth perception) any sort of factor in this ability to assign significance? (or are we just talking 2D images here?)

---

A colleague working in this field asked the same question a couple of years ago, and the answer was "We've never tried that!"This wasn't strictly true - there were attempts to do binocular anlysis on blockworld models. But in general, I think it was assumed that the 2D model must be easier to work with. There is some work being done now which tries to get additional information from 3D information. I'm going to try to find some references on this stuff so you can follow up if you like. I have colleagues who work in this field, but my own work is only tangentially related.

Interesting I would have thought that depth perception would be an advantage in learning how to discern significance ie looking at a dog in 3D it is easier to tell were the dog ends and the scenary begins as opposed to a 2D image. It seems to me that as humans start with binocular vision and learn to see in 3D where objects are more easily distinguished as distinct it seems that we would have an advantage in 2D.If this is true then a computer with binocular input of 3D images as part of the learning would provide a somewhat more level playing field....