Is consciousness necessary for intelligence?

[HarryW]

Algorithms are (as you know) a series of specific steps towards achieving a goal that will always achieve that goal. KBSs and similar AI-related things use heuristics, which are like rules of thumb. They work most of the time, but not all the time. They are useful when an algorithm is impractical. For instance:

Goal: prevent hijackers from boarding a plane.

Algorithm:
- search the plane
- strip search all passengers and crew as they board
- search all luggage before putting it on the plane

Heuristic:
- question and possibly search any passengers that match a predefined hijacker profile
- put passengers through a metal detector
- X-Ray all luggage.

The algorithm is totally impractical, but is guaranteed to work. The heuristic is practical, but is not guaranteed to work, although it will work almost all the time.

This is closer to what we use ourselves to make decisions.

[kedaman]

There's more important goal involved in the heuristic.
In the heuristic version it there seems to be higher priority in:
* The passengers convenience
* Time
Are not heuristic just algoritms which are aiming at a more "floating" goals. There's the one big goal in giving the company most profits, but the heuristic is part of the process, in a big algoritm so to speak.

[Nucleus]

Originally posted by HarryW
Perhaps? Is there or isn't there?

I would allocate to whatever seemed to me like the most likely to be fruitful.

On what basis do you determine which theory is going to be more fruitful?

Why would I think it was a good idea if there was no positive evidence and 'perhaps' even contradicting evidence? It doesn't make sense to me.

Subjectivity is defined as "subjectivity n : judgment based on individual personal impressions and feelings and opinions rather than external facts" Source: WordNet ® 1.6, © 1997 Princeton University

It is clear therefore that subjective opinion is not based on external facts. So it is perfectly valid to subjectively like a theory despite contradicting scientific evidence.

The hypothetical example I gave you asks you to choose between allocating funds on a subjective basis as opposed to adopting a more objective perpective. Hopefully the question makes more sense now?

[Nucleus]

Kedaman,

I decided look up the definition of evolution,

evolution n 1: a process in which something passes by degrees to a more advanced or mature stage; "the development of his ideas took many years"; "the evolution of Greek civilization"; "the slow development of her skill as a writer" [syn: development] 2: the sequence of events involved in the evolutionary development of a species or taxonomic group of organisms [syn: phylogeny, phylogenesis]


Source: WordNet ® 1.6, © 1997 Princeton University

Evolution encapsulates the concept of development. I don't believe random change can be considered evolution, as truly random change will not lead to development as time approaches infinity.

For example take Deep Blue. If the application was able to modify itself to improve its ability to play chess over time, then I would consider that to display an evolving interpretation and thus would label the application intelligent.

If on the other hand the application just randomly modified itself, then over time the computer would not be able to develop its ability to play chess and therefore would not consider the application intelligent.

[Nucleus]

Simon and Harry,

I don't fully understand your prescription for allocating funding on a subjective basis, and in particular how it may apply to the hypothetical situation I posed?

[simonm]

Nucleus

I don't fully understand your prescription for allocating funding on a subjective basis, and in particular how it may apply to the hypothetical situation I posed?

OK, in an attempt to answer your hypothetical situation, I would ask: "Why do I think an existing theory, that has some observation evidence, is a load of bollocks?". I could only conceive of this situation if the established theory provided a bad explanation. I might prefer a lesser known theory, despite it having less observational evidence, because it provided a better explanation.

So why should one explanation be considered "bad" and another "good"? Probably because one is more generalised or deeper than the other. One might be less dependant upon context than the other. One might just be a more convoluted way of describing the other. One might be unneccessarilly complex than the other. The list could go on but surely you get the idea?

[simonm]

Nucleus

I don't fully understand your prescription for allocating funding on a subjective basis, and in particular how it may apply to the hypothetical situation I posed?

OK, in an attempt to answer your hypothetical situation, I would ask: "Why do I think an existing theory, that has some observation evidence, is a load of bollocks?". I could only conceive of this situation if the established theory provided a bad explanation. I might prefer a lesser known theory, despite it having less observational evidence, because it provided a better explanation.

So why should one explanation be considered "bad" and another "good"? Probably because one is more generalised or deeper than the other. One might be less dependant upon context than the other. One might just be a more convoluted way of describing the other. One might be unneccessarilly complex than the other. The list could go on but surely you get the idea?

If you don't agree with a subjective basis for an allocation of funding, what do you propose? Is there a way that funding could be allocated objectively? Tell us what you think.

[Nucleus]

"Why do I think an existing theory, that has some observation evidence, is a load of bollocks?"

I answered this for Harry a couple of posts ago. By definition a subjective opinion is not based on external facts.

I advocate basing one's decision on objective facts, so I would apply funding to the theory with scientific evidence in support and go against my subjective opinion.

I do not feel anything constructive will come out of discussing this further. Let's just agree to disagree.

[HarryW]

If you want to get technical about it, with definitions and everything, nothing is purely internal. All our opinions are hugely affected by our environment. It is impossible to make a purely subjective decision, only mainly subjective. That's what I think people are talking about here.

[simonm]

Nucleus

I advocate basing one's decision on objective facts, so I would apply funding to the theory with scientific evidence in support and go against my subjective opinion.

What objective facts are you talking about? If we can't allocate research funding until we have demonstrated that there is supporting evidense for a theory, how are we supposed to get this "supporting evidence"?

Or would you propose an inductive approach? If general observations have been made that appear to support a theory then it deserves research? If you are, just say so and I'm sure someone will explain the flaws in inductive reasoning.

You also seem to be suggesting that we have access to objective facts about reality. we do, but only through a process of subjective interpretation.

If you do not wish to discuss it further then fair enough but I don't see how this converstation has nothing more productive to say (Unless you don't have anything new to add).

[csammis]

If you want to get technical about it, with definitions and everything, nothing is purely internal. All our opinions are hugely affected by our environment. It is impossible to make a purely subjective decision, only mainly subjective. That's what I think people are talking about here.

HarryW, how are you defining "environment"? If you define it as purely external stimuli, then what you have said is wrong. If I decide that I have to use the restroom, it is probably not because a Viking warrior came up to me and shouted Vikingese at me to the effect of "Go to the restroom!" More than likely it is because of my subjective decision that using the restroom would be beneficial for my comfort.

[kedaman]

Originally posted by HarryW
Nucleus, just read Simon's post for my reply

Kedaman, the point in a heuristic is that it's practical. The fact that the specific heuristic for the hijackers problem was condusive to the comfort of the passengers and the saving of time and money is a result of its need to be practical.

In general, heuristics are based on easily achievable, probable success, and are used in situations where algorithms for the same goal would be difficult to achieve.

Practical? I don't quite see your point as if there has to be a difference. If an algoritm is efficient, it fullfills it's purpose, if a heuristic is practical, it's efficient right? The algoritm version would be as practical, if the concern was to get the job done without concern about the passengers and time, and ultimately the money. You've set different goals in your samples, is it the goals that matters or is it the process?

Nucleus
Did you look up developement too?
The dictionary will hardly give you anything but practical definitions for everyday situations. If you find something like "evolution is the developement of something due to survival of the fittest, or natural selection" then you've found the right thing. AI that inherits this idea, called genetic algoritms performs evolution for a specific purpose, however unless the purpose is meaningful the prosess is useless. The survival of the fittest is a natural phenomenon, a result of some basic axioms. It doesn't imply a purpose.
Evolution is a problem though, Developement defines a subjective phenomenon, how can you distinguish developement without dealing with a purpose. If there's a goal something is developing towards, then it is developing in aspect of that goal, in aspect of another goal, it might not develope at all.

For intelligence to be a usefull concept it's definitions shouldn't contain any subjective contents.

[Nucleus]

Simon

What objective facts are you talking about?

E.g. the presence of any supporting scientific evidence.

If we can't allocate research funding until we have demonstrated that there is supporting evidense for a theory, how are we supposed to get this "supporting evidence"?

Well in the absence of any supporting scientific evidence, I would look at the pedigree of the person proposing the theory. If they are considered an expert in their area with many other well respected theories to their credit, then the probability that the proposed theory may be successful is increased. I consider this a more appripriate way of justifying my funding allocation rather than allocating funds based on intuition or feelings.

I guess I am advocating a rather conservative approach to resource funding, however, at the same time I do so in an attempt to maximise the probability and the value of each funding dollar.

If you do not wish to discuss it further then fair enough but I don't see how this converstation has nothing more productive to say (Unless you don't have anything new to add).

Ok, lets see where this goes then.

[Nucleus]

Kedaman

Did you look up developement too?
The dictionary will hardly give you anything but practical definitions for everyday situations. If you find something like "evolution is the developement of something due to survival of the fittest, or natural selection" then you've found the right thing. AI that inherits this idea, called genetic algoritms performs evolution for a specific purpose, however unless the purpose is meaningful the prosess is useless. The survival of the fittest is a natural phenomenon, a result of some basic axioms. It doesn't imply a purpose.
Evolution is a problem though, Developement defines a subjective phenomenon, how can you distinguish developement without dealing with a purpose. If there's a goal something is developing towards, then it is developing in aspect of that goal, in aspect of another goal, it might not develope at all.

For intelligence to be a usefull concept it's definitions shouldn't contain any subjective contents.

I guess as long as the application is able to improve an aspect of its interpretation of the world around it on its own, then I would consider the application intelligent. The application does not require a purpose for it to be deemed intelligent. I was just using a chess application to illustrate the concept of intelligence and that random change is not enough. If Deep Blue was able to improve its interpretation of another aspect of the world unrelated to chess, this too would be intelligence.

[kedaman]

Nucleus

improve an aspect of its interpretation of the world around it on its own

that's quite a lot subjective terms. We're not going anywhere with them.

[Nucleus]

Kedaman

By evolve I mean develop/improve/progess, etc well you get the idea. This can be objectively measured. For example a student that learns the rules of chess, but after a period of time is then able to beat his/her teacher is an example of an evolved ability to play chess. Computers as they stand today have a static ability to interpret the world and their ability to play chess is objectively fixed. Endowing a computer with a random ability to change its interpretation, cannot result in an improvement in interpretation and is therefore not intelligence. Applied to chess the computer would never get better at chess as time -> infinity. A computer with the ability to objectively and measureably develop its chess playing ability is by my definition intelligent.

Importantly the process by which it arrives at the improved interpretation does not have to be restricted to natural selection or any other traditional view of evolution. It is enough that there is an objective improvement in its interpretation of data achieved on its own.

I am confident you understand the concept of intelligence I wish to convey. If you can metamorphize the definition into one that you are happier with, by all means go ahead.

[simonm]

I guess I am advocating a rather conservative approach to resource funding, however, at the same time I do so in an attempt to maximise the probability and the value of each funding dollar.

Yes, indeed, it is a more conservative aproach to allocating research funding. Incidentally, do you actually have to allocate research funding?

What you require, really, is almost for a scientific theory to prove it's self worthy before you would consider it suitable for funding.
Whilst, as you said, this probably gains maximum value for the "tax dollar" spent on research funding, it doesn't ensure maximum progress as new ideas will inevitably come slower due to many good ideas being ignored.

I don't think I can really argue against that (as much as I might like to) as it's all down to a matter of how much you might have to spend on funding research anyway.

I still think you undervalue good explanations as a way of inspiring new directions in research, but there you go.

[HarryW]

Practical? I don't quite see your point as if there has to be a difference. If an algoritm is efficient, it fullfills it's purpose, if a heuristic is practical, it's efficient right? The algoritm version would be as practical, if the concern was to get the job done without concern about the passengers and time, and ultimately the money. You've set different goals in your samples, is it the goals that matters or is it the process?

An algorithm and a heiristic are fundamentally different. In some ways, algorithms are a kind of heuristic; it would be possible to make an algorithmic heuristic, but not all heuristics are algorithms.

Algorithms guarantee a result, heuristics don't. You couldn't make an algorithm that was practical in this situation, because an algorithm, by definition, must guarantee success, all the time.

If you want to think of them as similar by all means do because they are similar Heuristics have much more scope to them than algorithms though, because they aren't restricted to a guaranteed result.

Think about what you do in everyday life, for instance drive to work. The objective here is to get to work in time without crashing. An algorithmic solution for this would have to guarantee both these goals would be satisfied, and I can't think of a way to guarantee them both. In real life we just go for a heuristic solution - we're careful not to crash but we drive at a pace that will get us there on time. Neither goal is guaranteed to be satisfied but it is likely.

[kedaman]

Nucleus
I have to disagree it can be objectively measured. Progress is implying a goal, and a goal is the context for progress subjective perspective. However we can redefine intelligence as a function, based on the goal for progress. This can be acceptable after all.

#define Intelligent(x,y) (bool)(CanProcessInfo((subject)x) && Progresstowards((subject)x,(goal)y))

This is a illustration with a C++ macro to explicitely define intelligence for clarity. If there's something you don't get, ask.


Harry

So Algoritms are a special case of Heuristic which guarantee a result. So if I draw a success graph

A   B        C
     _________ Success
    /
   /
  /
_/             Failure

The whole range from A to C are Heuristics while the heuristics from B to C are Algoritms, am I correct?

To avoid subjectivity again, In such case a heuristic explicitely needs a goal. A heuristic with identical contents with another but has the slightest different set of goals are different by definition, Correct?

[simonm]

Kedaman

This is a illustration with a C++ macro to explicitely define intelligence for clarity. If there's something you don't get, ask.

Could you give a VB example please? I am not familiar enough with C++ to understand your definition.

[kedaman]

Function Intelligent(x as subject,y as goal) as boolean
   Intelligent=CanProcessInfo(x) and Progresstowards(x,y)
End function

[Guv]

Nucleus: You seem to object ot any subjective decisons. Consider the following, which I consider reasonable.

Suppose you have authority to allocate research funds. A well dressed person who seems bright shows you credentials indicating that he has degrees from MIT in astrophysics, mathematics, and chemistry. He shows you a 50 page proposal titled "Planets in Andomeda Galaxy made of various kinds of cheese."

Would you read his proposal? Would you try to verify his MIT credentials? If his credentials could be verified, would you consider his proposal more valid than if his credentials were phoney? If you read his proposal and it claims that observational evidence based on astral projection exists, would you allocate funds? Would you do any further objective analysis of this proposal?

As soon as I saw the title, I would decide to not fund the research, based on internal subjective analysis. Id Est: This proposal would immediately be put at the bottom of my subjective believabilty scale. I would tell the gentlman that I will read your proposal later and contact you with my decision, intending to never read the proposal.

I would spend my time doing an objective analysis of other proposals or chasing girls or watching TV or arguing with you folks on this forum. I would consider it a waste of my time to attempt to gather more data or apply any objective analysis to this proposal.

Other theories, less ridiculous that the above, would be treated similarly.

[Nucleus]

[Guv

You seem to object ot any subjective decisons

Yes when allocating funding I would prefer to base my decision primarily on external facts in an attempt to increase the probability that the research is successful.

Consider the following, which I consider reasonable.

Suppose you have authority to allocate research funds. A well dressed person who seems bright shows you credentials indicating that he has degrees from MIT in astrophysics, mathematics, and chemistry. He shows you a 50 page proposal titled "Planets in Andomeda Galaxy made of various kinds of cheese."

Heh heh. This is a joke right you don't honestly consider this example plausible do you?

I would check the calendar to see if it was April 1 and laugh heartily with them.

Would you read his proposal? Would you try to verify his MIT credentials? If his credentials could be verified, would you consider his proposal more valid than if his credentials were phoney?

The person's credential would make a difference. If the person lacked credentials I would politely ask him/her to leave.

If on the other hand he/she turned out to be a well-regarded genius in astrophysics with countless accepted theories to his/her credit, then I wouldn't dismiss the proposal as quickly. In this case I would at least turn over to read the synopsis. Perhaps this is evidence of a sense of humor, and a valuable new theory was just waiting for me to look past the front cover?

If you read his proposal and it claims that observational evidence based on astral projection exists, would you allocate funds? Would you do any further objective analysis of this proposal?

Assuming the person's credentials are supported and I have ascertained they are serious about the theory, based on the fact that no rational scientist would put forward astral projection as scientific evidence, I would be thinking that perhaps this genius has slipped quietly into complete madness. Before going any further I would request a full psychiatric evaluation.

Thanks for livening up the debate Guv!

[Guv]

Nucleus: Do you really think every proposed idea is worth amlyzing for possible evidence in its favor?

Is there nothing you would dismiss as not worth investigating?

If you were allocating funds for research, would you really investigate every proposal? Have you ever heard of comcepts like the biggest bang for a buck?

Do considerrations of cost effectiveness ever influence you?

If the crack pot community ever discovered some one like you was in charge of allocating research funds, the poor guy would be buried under tons of weird proposals. Hell, I would try for some of that money myself and I could come up with something more plausible than astral projection!

Going back quite a few posts, did you notice the areas in which I suggested that intelligent people would have differing belief systems? Like the frontiers of human knowledge and areas in which either or neither had much knowledge?

You posted something which suggested that you ignored that part of a post of mine.

[Nucleus]

Guv

Do you really think every proposed idea is worth amlyzing for possible evidence in its favor?

Is there nothing you would dismiss as not worth investigating?

If you were allocating funds for research, would you really investigate every proposal?

I think I made it quite clear that "Planets in Andomeda Galaxy made of various kinds of cheese." would be dismissed without a second thought if presented by someone without any credentials.

Do considerrations of cost effectiveness ever influence you?

Yes by choosing factors such as prior scientific evidence and pedigree I hope to increase the probability that funding will not be wasted.

[/quote]If the crack pot community ever discovered some one like you was in charge of allocating research funds, the poor guy would be buried under tons of weird proposals. Hell, I would try for some of that money myself and I could come up with something more plausible than astral projection! [/quote]

On the contrary, if you base your funding on your subjective judgment, all the crackpots have to do is work out how to influence your thinking, and voila instant cash. Only here does personal opinion triumph over reason and fact, consequently you would be their favourite benefactor.

Going back quite a few posts, did you notice the areas in which I suggested that intelligent people would have differing belief systems? Like the frontiers of human knowledge and areas in which either or neither had much knowledge?

Yes I vaguely remember... What was the point again?

[Bonker Gudd]

You are insane

[yrwyddfa]

Heretics. The lot of you.

[Shaggy Hiker]

Oh boy, someone dug up a fossil!!

[yrwyddfa]

Are you buried from the navel down, then?

[csammis]

Wow, I forgot that I even belonged to this forum until I got the subscription notification on this thread...but I sure remember this thread

In the three years since I had this debate, I've taken artificial life courses, neural networking courses, and a philosophy of technology course (I really wish I had this thread handy for that one). It amazes how much this topic actually influenced my way of thinking as I went on with computer science

[alkatran]

Don't neural nets simulate neurons?

Did you know a neuron either fires or it doesn't? That means there is no "maybe" at the most basic scale of humans OR computers. There is a cut-off point between true and false.

Genetic algorythms and neural nets are capable of doing VERY complexe things (image recognition, speech recognition, LEARNING to an extent). What happens when we have a neural net with same basic properties as you're neuron-net?

[KodeShark]

i woulda voted....but I was too lazy to read the options....

[dglienna]

I've heard of people getting grant money for some *really* ridiculous things. I'm amazed, and often thought of asking for some $$$.

[yrwyddfa]

Did you know a neuron either fires or it doesn't?

That's not entirely true . . .

Neurons fire like a 'leaky integrator' with a nice attack and a gentler decay on it's waveform. THus you get much more complicated neural firing patterns than you would with a Pitts-McCulloch model,

[yrwyddfa]

Option Explicit
 
Public Const e As Single = 2.71828182845904
 
Public Const ALPHA As Single = 25 'Rate of Change Of Activation:Decay Rate
Public Const BETA As Single = 20  'Rate Of Change Of Activation:Rise-Time
 
Public Function Sigma(a As Single, rho As Single, omega As Single) As Single
 
    'logistic sigmoid
    Sigma = 1 / (1 + (e ^ ((a - omega) / rho)))
    
End Function
 
Public Function LeakyIntegrator(Level As Single, SystemIn As Single, DecayRate As Single, RiseRate As Single) As Single
 
    'rate of change of activation
    LeakyIntegrator = -(DecayRate * Level) + (RiseRate * SystemIn)
    
End Function

Here's the two functions I use for my neural networks (straight from a text-book)

The logistic sugmoid function keeps the value between 0 and 1 (and flattens the curve a little) and the LeakyIntegrator ensures a more realistic waveform rather than a square sawtooth one.

[csammis]

Originally posted by alkatran
Did you know a neuron either fires or it doesn't? That means there is no "maybe" at the most basic scale of humans OR computers. There is a cut-off point between true and false.

This is definitely not true for biological neurons. Neurons can fire at different intensities, and they have thresholds for firing on differing levels of input or different combinations of input.

As yrwyddfa pointed out, nodes in a neural net most often use values between 0 and 1. This is due to the concept that neural nets are probability-based difference engines, as are some models of the human brain