Is Biological Intelligence Different From Machine Intelligence?
#13
(11-Aug-2011, 05:16 PM)sojourner Wrote:
Quote:Sweeping dismissals of the use of a computational metaphor are unwarranted

Is the dismissal of a proposal of a god as a bank of supercomputers unwarranted?

I do not know much about behaviorism, nor about cognitive science. But I see Arvind's response as addressing this claim of yours:

(11-Aug-2011, 03:34 AM)sojourner Wrote: My point is that work done under cognitive science will be of zero benefit for working with humans.

He gave some specific examples (Works of Tversky and Kahneman, and Antonio Damasio). I don't see how you can dismiss the computational metaphor as being the same as god without addressing those examples.
Reply
#14
Scott Aaronson is not arguing by analogy. He is making a direct claim. It is not about 'comparing' the biology of the human brain with the concept of computational complexity, which would be absurd considering that 'computational complexity' is a mathematical concept used to characterize at an abstract level, tasks such as problem solving. For instance, one of the first problems one tries to solve in computational complexity is suppose I have an unsorted sequence 1 to n, I wish to understand as a function of n, how many fundamental 'atomic' operations will it take me to sort them. The 'fastest' way one can do this is with operations that increase as a linear function of 'n'. The point he makes about a computer learning all the conversations of humans and using it as a database is that it suffers from the problem of curse of dimensionality. It may happen, no matter how advanced the technology, the lower limit for the time taken to compute on such a database may be absurdly, astronomically, large.

In this he contradicts the slightly mystical claim by Roger Penrose, who quoting from a quote from Aaranson's paper said:

Quote: Roger Penrose said,
Quote:One could equally well envisage computers that contain nothing but lists of totally false mathematical ‘theorems,’ or lists containing random jumbles of truths and falsehoods. How are we to tell which computer to trust? The arguments that I am trying to make here do not say that an effective simulation of the output of conscious human activity (here mathematics) is impossible, since purely by chance the computer might ‘happen’ to get it right—even without any understanding whatsoever. But the odds against this are absurdly enormous, and the issues that are being addressed here, namely how one decides which mathematical statements are true and which are false, are not even being touched...

Aaranson claims, this slightly mystical approach is not necessarily the right one. One should base the solution on the computational complexity of such a task.

To quote Aaranson, (in parenthesis are my remarks)

Quote:The trouble with this (Penrose's) response (to the question of whether a computer can simulate a brain) is that it amounts to a retreat from the sword-in-the-stone test, back to murkier internal criteria. If, in the end, we are going to have to look inside the computer anyway to determine whether it truly “understands” its answers, then why not dispense with computability theory from the beginning? For computability theory only addresses whether or not Turing machines exist to solve various problems, and we have already seen that that is not the relevant issue. To my mind, there is one obvious direction that Penrose could take from this point to avoid incoherence—though disappointingly, it is not the direction he chooses. Namely, he could point out that, while the lookup table “works,” it requires computational resources that grow exponentially with the length of the conversation!

The paper posits this:

Quote:(*) Any computer program that passed the Turing Test would need to be exponentiallyinefficient in the length of the test—as measured in time, memory usage, or simply the number of bits needed to write the program down. In other words, the astronomical lookup table is essentially the best one can do.

This is wholly consistent with a rationalistic view of human intelligence. A question to ponder from a meta-perspective is this: is there an 'unknown' feature of the human brain that makes it uncomputable? This is highly unlikely from an observation of human behavior. Granted, we are immensely complex. But does this mean that the human brain cannot be simulated using technology?

Aaranson says,

Quote:However, (computational) complexity theory (theoretical computer science) also makes it clear that, even if we supposed (*) held, there would be little hope of proving it in our current state of mathematical knowledge. After all, we cannot even prove plausible, well-defined conjectures such as P != NP.




"Science is interesting. If you don't agree, f off." GoodMorning
[+] 1 user Likes karatalaamalaka's post
Reply
#15
A quick reply now. I will post a more elaborate reply at a later date, if necessary.

Quote:Tversky and Kahneman, and Antonio Damasio

I am not familiar with the work of either (group) of the above. That's why I didn't say anything about them. I plan to look at their work.

However, it can be safely said that cognitive psychologists (including the above) are NOT interested in discovering basic processes that bring about behavior and learning in the individual animal. Doing so necessarily involves paying attention to the stimulating conditions and consequences, to which cognitive psychologists pay only scant attention if any.

I don't see how computing plays any part whatsoever in the molecular responding and learning of individual animals.

A part of human behavior is verbal -- both public and private. This is also affected by the usual behavior analytic variables of stimulus control, motivation, emotion, and consequences. I don't see how task or algorithm comes in in the biological molecular responding and learning of individual animals. Searle seems to be dead on.

The basic principles discovered by behavior analysis are applied in various areas -- such as autism treatment and getting someone to say "pencil" at a certain time (the latter for demonstration purposes).

Can cognitive science's tasks and algorithms contribute anything today towards helping in the treatment of autistic children?

Can cognitive science's tasks and algorithms contribute anything today in getting someone to say "pencil" under a given set of conditions? [We could choose the name of any other object than "pencil". I am staying with this because this is the one in the classic book.]






















Reply
#16
I am posting this as a separate post because I don't want the arguments here to be confused with the ones in my immediately previous post.

Simulating the brain: Isn't the goal to simulate function rather than structure? If so shouldn't we talk in terms of simulating behavior (including changes in behavior called learning)? Also, in order to do so, shouldn't we take advantage of principles already discovered by fields such as behavior analysis? I think a certain thought at a said time because of

(a) my conditioning history

(b) recent/current stimulation

© current deprivation/satiation conditions (called motivation)

(d) current emotional conditions (triggered by stimuli)

(e) the current value of the Dow Jones average [this is a joke :-)]
--------------------------------------
It seems to be that two distinct topics are mixed up in the current thread (which is already a split one):

Getting machines to exhibit human like behavior

Understanding human behavior -- is the computer analogy of any use in this?

Is a further split necessary?


Reply
#17
(12-Aug-2011, 04:52 PM)sojourner Wrote: A quick reply now. I will post a more elaborate reply at a later date, if necessary.

Quote:Tversky and Kahneman, and Antonio Damasio

I am not familiar with the work of either (group) of the above. That's why I didn't say anything about them. I plan to look at their work.

However, it can be safely said that cognitive psychologists (including the above) are NOT interested in discovering basic processes that bring about behavior and learning in the individual animal. Doing so necessarily involves paying attention to the stimulating conditions and consequences, to which cognitive psychologists pay only scant attention if any.

But your earlier point was "..work done under cognitive science will be of zero benefit for working with humans." My point is you can't outright dismiss cognitive science without first showing that it is of no use or benefit to humans, to which the examples given by Arvind stand as a counter point.
Reply
#18
Perhaps what I write here on this topic should be accompanied by the disclaimer that computational approaches to cognitive science are not entirely unrelated to my own academic specialization, due to which I may depart from what is considered a 'settled' historical narrative often.

Cognitive Science today is 'Darwinian' in that our sensory and decision-making apparatus are studied as having their 'default settings' optimally adapted to the statistics of frequently encountered stimuli in a given ecological niche. In principle, this goes without saying, but in practice it begs the question of what defines 'evolutionary fitness' and what quantity it is that Evolution is trying to optimize for (eg. Does our visual apparatus optimally trade off detection speed with accuracy?). Natural-selection-based narratives applied either at the level of behaviors themselves or at the level of neural networks are quite mainstream. Beginning as it did as a subdiscipline of Biology, it is scarcely surprising that Cognitive Neuroscience employs a Darwinian idiom.

Cognitive Science today can also be thought of as conforming to 'behaviorism' in that the emphasis is on 'cognitive processes' (eg. scene parsing, language processing, decision-making under risk) rather than a study of the phenomenon of 'personality'. Any sufficiently realistic model of scene parsing, language processing or decision-making under risk must incorporate some uncertainty and hence Cognitive Science today is also 'probabilistic' in the language it employs. The computational resources for widespread deployment of the workhorse of probabilistic reasoning,namely, Bayesian inference became available only since the 1980s. In the past few decades has become an indispensable part of the Cognitive Scientists' toolkit. It is worth noting here that the word 'computational' itself has undergone a change in connotation in the past few decades from an emphasis on formal logic towards more probabilistic reasoning. Quoting mathematician David Mumford from his 2000 paper "The Dawning of the Age of Stochasticity":

Quote:“Thought is the weighing of relative likelihoods of possible events and the act of sampling from the ‘posterior’, the probability distribution on unknown events, given the sum total of our knowledge of past events and the present context. If this is so, then the paradigmatic mental object is not a proposition, standing in all its eternal glory with its truth value emblazoned on its chest, but the random variable x, its value subject to probabilities, but still not fixed.”

Cognitive Science today is also becoming increasingly 'empirical' given the advent of measurement modalities like functional magnetic resonance imaging in the 1990s. While saying that it is 'empirical' one must acknowledge that secure theoretical foundations are pending and to quote V S Ramachandran, the cognitive sciences are in a 'Faraday stage rather than a Maxwell stage'. Cognitive Science therefore today draws routinely upon such techniques as Bayesian inference and such technologies as fMRI, both of which entered the mainstream in psychology and neuroscience well after B F Skinner's retirement in 1974.

In sum, Cognitive Science today has a character that is Darwinian, behaviorist, probabilistic and empirical due to a historical confluence of several breakthroughs and trends and hence deserving of a fuller narrative rather than centered around a single investigator, no matter how towering. Eponymous narratives have their utility in science popularization and science biographies written by James Gleick or Walter Isaacson remain worthy additions to a science lover's bookshelf, but these remain supplements rather than the substance of scientific study.






[+] 2 users Like arvindiyer's post
Reply
#19
This is with reference to the following quote from Sojourner's latest post:

Quote:Traditional psychology assumes that there is a non-physical entity that resides inside us that is responsible for our thoughts and actions - this entity is called our mind/soul/spirit and the like. Some people can let go of this, only to replace it with a computer inside us that they talk about using the analogy of artificial intelligence.

As for a 'philosophy of science'-based treatment of the ongoing modern synthesis in Cognitive Science, the view is entirely materialistic. The abstractions employed in the computational metaphor can by no means be considered on par with the vacuous abstractions of vitalism, because these computations are eventually rendered explicable as phenomena involving physical entities, namely neurons. It is true that 'Good Old-Fashioned Artificial Intelligence' did not concern itself with the neural implementations of the computations it was studied (and in this sense the above quote is not way off the mark), but the discipline of Computational Neuroscience definitely does in the 'third Marr level' as explained in an earlier post. This reducibility of cognitive processes to neural phenomena obviates any 'homunculus' myths, rejects Cartesian Dualism and according to philosophers of science like Patricia Churchland, are best addressed in a philosophical framework of Eliminative Materialism. To say that Computational Neuroscience of today is a homunculus in disguise would be not just erroneous but a gross mischaracterization. A useful resource in this regard is the text The Computational Brain by Churchland and Sejnowski.
[+] 1 user Likes arvindiyer's post
Reply
#20
Despite the fact that biological machines evolved over billions of years while what we humans generally refer to as machines are those that were (usually) intentionally designed by us using non-biological materials, with an understanding of the mechanistic functioning of these objects, neither group is comprised of non-physical elements in the naturalistic sense.

Time and imagination limit non-biological machines from attaining qualities akin to those we possess.

The universe has the effect of an immensely powerful computational system and natural selection (along with certain other non-adaptive evolutionary processes that led to biological life) is an immensely complex, malleable and open-ended "evolutionary algorithm" (used in a qualified sense) involving, from our point of view, infinite parameters and trajectories acting over hundreds of billions of evolutionary units over hundreds of millions of generational cycles. In other words, biological intelligence was "designed" by a process that is infinitely more powerful, involving infinitely more computational ingenuity, than all of human endeavour since the invention of fire put together. So, of course in most respects we are like 'god' machines in comparison to the simple machines we design.

If we are talking only about machines created by us so far, then of course, YES, biological intelligence is vastly different from the most brilliant computers (or any other machines) we have. But if we are making a general statement about the potential for intelligent biological systems to 'create' anything like biological intelligence using non-biological materials given enormous amounts of time and resources , I say it is not impossible.
"Fossil rabbits in the Precambrian"
~ J.B.S.Haldane, on being asked to falsify evolution.
Reply
#21
Saturday is my day for running errands, which I do not do very well :-) I plan to post on this thread today between errands.

Quote:[Skinner was] accused of (and even vilified for) suggesting that people be viewed (as being quite malleable)

The above is true. It has been said that Skinner has had the worst press of anyone since Darwin. And views attributed to him are often based on second-hand and third hand sources.

For instance, Skinner never viewed humans as automatons.

1. In his worldview, genetic contribution to behavior plays an important role. He pointed out multiple times that operant conditionability itself is of genetic origin. In the evolutionary chain, it appears at some point near spiders. (Source: A class I took.) I can't remember whether spiders have it or not.

2. His field was changing behavior during the lifetime of the animal. In his lab experiments, pigeons (I think) were carefully chosen based on their genetic abilities. He has made this very clear. The goal is to increase success to the maximum possible extent.

3. His students and later colleagues found and reported that behavior of genetic origin intrudes/overrides behavior conditioned using the law of effect.

My point: It is wrong to attribute a tabula rasa view to Skinner since that was never the case.
------------------------------
@Ajita:

My goal/interest is that of psychology, namely how to understand the behavior of humans so that we can produce useful technologies that benefit humans. My view is that the computer metaphor is totally useless in work towards this goal.

I don't believe that there is a vital force or anything like that in humans. It is all molecules. The human body maintains a structure within limits that makes various processes possible. This is all physical. I don't believe in a dualism at all. Nor do I believe in a mind.

Quote: biological intelligence was "designed" by a process that is infinitely more powerful, involving infinitely more computational ingenuity

I find the reference to computation wholly unnecessary. Can we not think of anything complex without introducing computation? I find Searle's comments spot on.
Reply
#22
I have some gladioli blooming in my living room. We talk about the plant's growth and blooming in wholly natural science terms without involving the computer in any way? Why the computer for describing human behavior?
Quote:Behaviorism is seen as integral to Psychology but not as it entirety and not entirely in the form intended by its original popularizer.

I have to ask such as what to both the above claims.

Have you studied/understood Skinner's views on verbal behavior and rule governed behavior? The latter term is used in a very specific way in behavior analysis.

Skinner proposed a natural science of behavior, a branch of biology. He also advanced it quite a bit in the sense that he discovered several basic processes (which have given rise to technologies in several fields). He also expected his discoveries to go out of date as is the case in any natural, empirical science, as more and more work is done. Some of this has happened (joint control). However, for the most part this has not happened. The right people don't take up psychology. They have to contend with constant hostility from the majority view. A lot of work needs to be done. However, I can't think of anything missing in psychology other than work to be done under behaviorism.
------------------------
I am assuming that the readers are interested in giving everybody a fair shake, including Skinner. Here is a link to a defense of Skinnerian views.

This is useful too because it is recent.

Quote:'Skinnerism' similar to the place Taylorism

This is the epitome of damned by faint praise and is like saying that Einstein took some physics courses :-)

Skinner's techniques are standard in behavioral pharmacology. They have revolutionized education though these advances are languishing. They are also the only working thing in autism treatment. There are about 5 major journals and probably about at least 10 minor ones. They are being practiced in at least a dozen countries.

And the book Verbal Behavior is a milestone in the history of thought. [I better leave home before the banks close :-)]


Reply
#23
(13-Aug-2011, 06:38 PM)sojourner Wrote: @Ajita:

My goal/interest is that of psychology, namely how to understand the behavior of humans so that we can produce useful technologies that benefit humans. My view is that the computer metaphor is totally useless in work towards this goal.

I was specifically addressing the question " Is Biological Intelligence Different Than Machine Intelligence?". I haven't given much thought to the question you are addressing.

Quote:I don't believe that there is a vital force or anything like that in humans.

I don't think the majority of us here do. I and most others here are naturalists/materialists.

Quote:
Quote: biological intelligence was "designed" by a process that is infinitely more powerful, involving infinitely more computational ingenuity

I find the reference to computation wholly unnecessary.

Since you've established you are concerned with a different question from the one I am trying to answer, I don't know why you think "reference to computation is wholly unnecessary". Different questions will necessarily focus on different aspects, even if addressing the same system. My focus in my exploration of the idea was on the processes that led to their being. That is what led me to say "The universe has the effect of an immensely powerful computational system and natural selection (along with certain other non-adaptive evolutionary processes that led to biological life) is an immensely complex, malleable and open-ended "evolutionary algorithm" (used in a qualified sense) involving, from our point of view, infinite parameters and trajectories acting over hundreds of billions of evolutionary units over hundreds of millions of generational cycles."

It seems to me you are not refuting my facts, but are dismissing their relevance, which is why I point you again to the fact that we seem to be concerned about different questions.


"Fossil rabbits in the Precambrian"
~ J.B.S.Haldane, on being asked to falsify evolution.
Reply
#24
(13-Aug-2011, 06:42 PM)sojourner Wrote: I have some gladioli blooming in my living room. We talk about the plant's growth and blooming in wholly natural science terms without involving the computer in any way? Why the computer for describing human behavior?

I don't think anyone here is claiming that we cannot talk about humans (or plants) unless we see them as computers (of sorts). But why should we NOT see them as computers if that is the focus of our question?

"Because that is not useful to us" is not a wholly satisfactory answer, for two reasons.

1. The question is often posed as an attempt to understand the system as in much of fundamental science, not necessarily to provide immediate practical applications.
2.The practical applications of various discoveries made in science are not realized until decades or even centuries later, often because they are not immediately apparent.

If a 19th century scientist were to talk about, say, photosynthesis in gladioli, s/he would probably describe the shape of the leaves, note the shape and number of stomata, measure transpiration rates, rates of oxygen release, carbon-di-oxide consumption, etc. A scientist in the late 20th century studying the same plant might be concerned about biochemical reactions in the photosynthetic pathway, allele frequencies that lead to differences in photosynthetic rates between genotypes of the species etc., in relation to questions that might not make sense to the scientist from the previous case, but still remaining firmly on the subject of photosynthesis. If the scientist in the 20th century wanted to ask the questions that the one in the 19th was concerned about, the methods used back in the 19th would suffice even in the 20th century. But the questions have changed. The questions that each scientist is concerned about are constrained by contemporaneous knowledge limitations.

In the 21st century, certain new questions are being asked that would not have been possible just a few decades ago- questions that require us to view the plant as a computational machine. In the context of photoshythesis, here is one such question that has been answered recently.
Quote:"Inside every spring leaf is a system capable of performing a speedy and efficient quantum computation, and therein lies the key to much of the energy on Earth."
Let's keep in mind, we, as a species, are just beginning to learn about computation. As I said before, our computation systems are extremely primitive in comparison to what the universe has achieved given its vastness in scale (in all dimensions). Our mathematics is also very primitive, and if we are to believe scientists like Wolfram, the explosion in new mathematics brought on by improvements in computation will result in a new scientific paradigm. Let's not even go that far. Let's begin by acknowledging the role that computational systems offer us today.

Since we are talking about plants, I'd like to point out that plant developmental biology is one area where computational methods are increasingly being used. The University of California at Irvine runs such a project. Again, when it comes to plants, these techniques are in the beginning stages, and even the questions are not fully coherent.

But the question we were originally concerned about is regarding 'intelligence' in humans and in machines. There is little doubt that a computational framework is directly relevant when talking about intelligent systems. Of course at the mechanistic level concepts such as life, intelligence etc. do not imply any non-physical qualities. But these are not meaningless concepts in scientific terms. There are emergent qualities associated with such phenomena as life and intelligence, that can be studied scientifically. And one way to study them is as computational systems.

This is not to say that good old behaviorism will be made redundant, any more than discovery of quantum computational effects in plant photosynthesis makes understanding the biochemistry of the process pointless. These are different levels of questions. The fundamental questions make it possible to ask the more complex ones. We are children in our understanding of computational complexity. I am not saying we will soon completely understand how the brain works by studying its computational mechanisms, but I think we can safely say that as our understanding of computational systems and the complex mathematics they are capable of increases, computers will become increasingly relevant in understanding how intelligent systems work.
"Fossil rabbits in the Precambrian"
~ J.B.S.Haldane, on being asked to falsify evolution.
[+] 1 user Likes Ajita Kamal's post
Reply




Users browsing this thread: 2 Guest(s)