On Intelligence...

個人檔案Klatu verata niktu相片部落格清單更多

工具

說明

部落格

排列方式：

20 July

My brilliant friend, David, recently posted an irresitable and deep blog entry... original article appears here. I liked this entry so much, I'm reproducing it here along with my comments just so those visitng my site could be encouraged to explore this interesting and thought-provoking issue for themselves.

I'm curious as to what you think about this issue. I'm sure David is too. Why not pay him a visit and let him know how much you appreciate his beautiful question?

David's Thought-Provoking Question

(Reproduced here with his permission)

Intelligence defined

Allow me to propose a new (as far as I know) definition of intelligence:

A system is intelligent to the extent that it can generate high quality choices and decisions.

It is always risky to attempt to redefine a widely-used term but I think it can be worthwhile if it leads to new insights while still capturing the connotations of previous definitions. Intelligence has been notoriously hard to define (it is an ongoing joke within the field of AI). But how can you hope to understand an intelligent system (let alone build one) if you can't even define it? I hope this new definition sheds some new light on some old questions: What kinds of entities are intelligent? How does one recognize or test for intelligence? Are there different kinds and domains of intelligence? Does it matter how something is intelligent, i.e. can something act intelligently but not actually be intelligent? How does consciousness figure in (if at all)? Can intelligence arise from non-intelligence? Is it an emergent behaviour? Are animals intelligent? Can a computer (program) be intelligent?

What do you think the answers are given the above definition?

My Response

As with any scientific pursuit, often the most difficult part of embarking on greater understanding is being able to frame the problem in an appropriate and precise manner.

I believe the proposed definition above achieves the first goal (appropriateness) but not the other (precision). The problem I have is with the subjective and value-judgmental term "high quality". Such subjectivity necessarily diminishes from the usefulness of the definition. Instead, I propose a minor change to remove the ambiguity and room for interpretation:

A system is intelligent to the extent that it can react to changes in its perceived environment in ways that decrease local entropy.

This definition removes the subjectivity of terms like "high" and "quality". Making order of disorder is arguably an excellent measure of intelligence. As an example, the directed, ordered firing of neurons produces thoughts, ideas, choices, decisions, and actions. No one can argue that the electrical signals of the brain differ little from the electricity in a rain cloud. However, the process within the rain cloud is entropic while the process in the mind is locally counter-entropic.

Another example, erosion destroys structure over time thus increasing entropy, but intelligent life creates order and structure through action, thus decreasing local entropy.

So, now we have an answer to the first question...How does one recognize or test for intelligence?
- Look for signs of decreased local entropy (increasing complexity and order in: structure, form, movement, action, behaviour, etc.) resulting from changes in environment.

Are there different kinds and domains of intelligence?
- Without a doubt, there are. A feeling is not the same as an idea, nor is it the same as an action. Each of these requires a kind of intelligence. The interplay of these near-infinitely variable qualities is the music to which our behaviours are choreographed.

- There are also different levels of intelligence arising from different levels of organization. None can argue that a bee is not intelligent nor fail to recognize that a single bee and a swarm of bees do not necessarily exhibit the same kind or level of intelligence.

Does it matter how something is intelligent, i.e. can something act intelligently but not actually be intelligent?
- A purely random process may over short observation windows exhibit entropy-decreasing effects, but the likelihood of sustainment of such effects decreases probabilistically with each passing unit of time, and such effects are not correlated to changes in environment.

How does consciousness figure in (if at all)?
- Consciousness is simply a product of higher intelligence. At some point of increasing organization and complexity an intelligence becomes self-aware. Self-awareness is born when an intelligent system suddenly recognizes that it is an intelligent system.

Can intelligence arise from non-intelligence?
- Of course... brains are made of neurons, which are living matter but arguably non-intelligent. As another example, thoughts are products of intelligence, and yet they are produced by simple (non-intelligent, non-living) electrical impulses in the brain.

Is it an emergent behaviour?
- Yes, I believe intelligence is an emergent behaviour. Of all things, I have to believe this if I believe any of the things I have written above. From basic electrical impulses we eventually get thought. How can this be if intelligence is not an emergent quality of an organized system? I believe that there is a critical stage of organization and complexity at which intelligence emerges as a quality of the system in question. At even higher stages of organization and complexity, consciousness emerges. Increasing complexity of organization and interaction produces higher levels of awareness and intelligence. Assuming for a moment, that all of this is true, then what can we expect to emerge beyond consciousness?

Just for fun, I propose the following unfounded conjecture:

What about a kind of hyperconsciousness? Nothing metaphysical or anything like that, but perhaps an intelligence so great as to allow one to assess and anticpate the actions of others before they enact them? Perhaps a predictable and accurate means of reading the other guy's next move by rapidly assessing the facts on hand and determining his next most probable course of action... like in a chess game but in real life?

Are animals intelligent?
- Of course!

Can a computer (program) be intelligent?
- Why not? As complexity and organization increases, how do we know that it will not reach the critical threshold at which intelligence emerges? And why not at even higher levels, self-awareness?

- Or put in another, more direct way... what if a computer fast enough and contained sufficient capacity existed to accurately model the connections and electrical characteristics of all the neurons of an average person's brain, and do so in realtime? Would this simulation not run at the same speed and simulate the same patterns of electrical impulses that occur in a fully functioning biological brain? If it produces the same "thoughts", why then would it not be deemed intelligent? How long will it be before such a computer exists?

9:04 | 發佈至您的部落格 | Personal

回應 (9)

若要新增回應，請以您的 Windows Live ID 登入 (若您使用 Hotmail、Messenger 或 Xbox LIVE，則您已擁有 Windows Live ID)。登入

沒有 Windows Live ID？註冊

	SprisslerCurtis撰寫： I think a better definition is that intelligence is the ability to produce desired results. Otherwise, you have to define "high-quality" too, you must have a way of telling a bad decision from a good one. And if you define it by decisions, where does that leave esoteric forms of intelligence like musical ability or memory? So it's an ability to produce desired results. In otherwords, intelligence is not independent of desire. (but not vice versa) Can a machine have desire? Sure. It just has to spontaneously care. It could develope desires in the same way humans have evolved to have them, or it could be given them by another entity. Can a tree have desire? No. And I doubt anyone would argue that it is intelligent. Are animals intelligent? Not as much as humans. And certainly not all of them. (you can have desire without intelligence as desire is a feeling) Can intelligence arise from non-intelligence? It pretty much has to, unless you're one of the "intelligent design" people. Can something act intelligently without intelligence? Of course it can, by accident, or by default. A hive of bees can make very good decisions - saving up food, etc - and it's not by virute of intelligence, but rather default: Bees that don't act this way die off and are replaced by those that do. (this is similar to the process by which we as a speciesattained intelligence) How do you recognize intelligence? It's fair to say that you must be intelligent in order to recognize it. You must have intellect in a particular category in order to compare actions to what you believe to be intelligent, assuming the same goal of outcome. Consciousness? Is not critical to intelligence, but seems to be directly proportional to it. Are there different domains of intelligence? Yes. I would break it into 3 parts: functional, actual and potential intelligence. Or did you mean categories like logic, musical aptitude, physical coordination, memory, etc? Can we measure intelligence? Absolutely (in the mathematical sense of the word), but we probably can't very well assess the intellience of anything smarter than we are. There are things we'd miss entirely. I don't believe IQ tests are good for anything but identifying the stupid. What about this question: Is a baby intelligent? It can't make decisions, but it can record and organize information in a rudimentary (often unconscious) way that eventually leads to an ability to make decisions. This is what I meant by potential intelligence. An entity that will in time become intelligent without being modified externally. And this: Does allowing desire and feeling to govern your decisions make you less intelligent? Less able to recognize intelligence? (in other words, can fear make you stupid?) I think so, certainly. Intelligence does nothing if you don't use it. Stupid is as stupid does. Ha ha ha.... But probably not less able to recognize it. This is the defining aspect of functional versus actual intelligence. A functionally stupid entity could actually possess great intelligence, and since you have to be intelligent to recognize intelligence, we could tell that their actual intelligence differs from their functional intelligence by observing their ability to recognize other intellect. 12 月 19 日
	-tymalone- 撰寫： hmm... yes, wow, quite interesting. i like your little unfounded conjecture. i would really like to provide you with a more in-depth comment on this in order to further extend the discussion, but i don't have the time. i had to say something though. anyway, keep thinking. Ty 11 月 29 日
	Zwixo 撰寫： Hi there, I read what you wrote, and i'm still amazed at the way your line of thought went. The vocabulary makes it like i'm on the highway with a bike! but on a serious note, i join neuronaut in his answers, i would also say intelligence is not what your talking about, organization: yes, science: yes, systematic: yes, thought: NO. And for me thought is the most intelligent feature in humans. And technology seems way far off right now. 7 月 21 日
	neuronaut42 撰寫： Say an intelligent agent (usually a human) is driving a car. They have to make several decisions every second: where to steer, how much to accelerate or decelerate, whether to turn the turning signal on or off, and where to look next. Do any of these choices increase or decrease entropy? A poor decision could lead to a crash and a large increase in entropy, but otherwise I would say entropy in unchanged for the most part. 7 月 21 日
	Søndis 撰寫： Did I just die and wake up in Harvard heaven??? This stuff is way more than I can even grasp at the moment, no one ever wrote in any brochure about Lebanon that there's 30 degrees at night.... I'm gonna sue that weather guy for leaving the oven on 24 hours a day... If you want to check out my own feeble attempt on artificial intelligence, check out http://www.building57.com and follow the link to "box motion". This is "female" and "male" boxes bumping into each other and creating new boxes. There's an age limit on each box, and each box can only reproduce within a certain age scope (like humans). After starting, it's just down to patience to se how long life will last under the given circumstances :) Have a good day, and thanks for keeping the planets brains ticking! Thomas 7 月 20 日
	neuronaut42 撰寫： I meant to say "I do agree it would be nice if intelligence could be measured objectively but I don't think your refinement works." 7 月 20 日
	neuronaut42 撰寫： My hyperbolic friend Simon made a valiant attempt to rescue my definition from the vaguaries of subjectivity. I do agree it would be nice if intelligence could be measured subjectively but I don't think your refinement works. The reason is that there are intelligent choices that increase entropy and unintelligent choices that reduce entropy. Though there may be a correlation between intelligence and reducing entropy, the fact that there are many counterexamples means that if we want to find an objective measure we have to look elsewhere. I skeptical that such a measure can be found. One of the most important implications of my definition is that it takes intelligence to recognize intelligence. Fortunately there seems to be a significant amount of agreement on what constitutes intelligent behaviour among intelligent people, at least in some domains which is why IQ tests can exist. But there are also domains where intelligent people disagree vehemently (politics comes immediately to mind). 7 月 20 日
	BigBluePlanetCorporation 撰寫： Aneen, good convo with your buddy there. THAT is what the internet is for -- sharing ideas in a many-to-many forum. I like that def., but my beef is with: 1) the response about Consciousness -- that is iffy and a circular argument, but very well could be true! AND 2)" first question...How does one recognize or test for intelligence? - Look for signs of decreased local entropy (increasing complexity and order in: blahblahblah, yaddayaddayadda...) I don't think that we seriously --- well let me paraphrase myself...Ok...actually, I changed my mind... I see where you're going. bad ass, you are bad ass ---- Actually you're a Bad Ash! And I'm Good Ash! I'm goodie little two-shoes, little goodie two-shoes! meegwetch, sg 7 月 20 日
	Nyniane22821 撰寫： WOW!!! Um.....WOW! I like your answers to Davids questions. I read them yesterday and my mind went crazy doing circles...now I see your answers....guess what? My mind is still doing circle. LOL! You are so smart....smrt. 7 月 20 日

引用通告

此內容的引用通告是：

http://kilmati.spaces.live.com/blog/cns!50E1454EFD60E60!180.trak

引述這則內容的部落格