An encyclopedia has a lot of knowledge. That doesn't make it intelligent.
The first definition for "intelligence" at reference.com is "the capacity to acquire and apply knowledge." That's an incomplete definition. There are programs out there that are able to "acquire and apply knowledge" (namely, chess expert systems) that I would not describe as "intelligent."
The second defintion, "the faculty of thought and reason" gets closer by adding the ability to "reason." However, to reason simply means to "think logically." Computers are very logical "thinkers," so, again, this doesn't quite hit the spot.
The third definition, "superior powers of mind," is too ethereal for our purposes.
So, what's the true definition of intelligence? Here it is:
intelligence - the ability to obtain and utilize knowledge in innovative ways
Innovative Use of Knowledge
The word "innovative" is key in our definition of intelligence. It is the innovative use of knowledge that makes an entity intelligent. In this context, "innovative" means "new" or "unprecedented," and is relative to the intelligent entity. For instance, using newly acquired knowledge to ride a bike is innovative for the individual, not the population.
Searching Google for "artificial intelligence" turns up a ton of links, some worth more than others. In particular is a set of pages authored by John McCarthy called "What is Artificial Intelligence?" I've used it as a template for my own ideas about artificial intelligence.
Only One Intelligence
Given our definition above, there is only one type of intelligence- the kind found in human beings. Intelligence is an all-or-nothing proposition. That's why I don't like the term "artificial" intelligence. There's never anything artificial about intelligence.
It's like being pregnant. You either are, or you aren't. There's nothing in-between. You can't put a pillow in your shirt and call it "artificial pregnancy."
You can fake intelligent behavior, however, which leads us to a few more definitions:
- true intelligence - an entity that is intelligent; the intelligence exhibited by human beings; not known in any other life form on planet earth; a full sentience of self-awareness and self-determinism (that is, free will). motivated by peace and joy.
- simulated intelligence - an entity that "acts" as though it is intelligent, but in reality is not. "Expert systems" are in this category. Follows pre-programmed patterns. Can never break free from programmed responses because of the lack of self-awareness. Lacks motivation and purpose. Can be a knowledge wonder, but can never be truly "intelligent."
Most people misuse the term "artificial intelligence" to refer to a "true intelligence" created by man. Therefore, I do not use the term "artificial intelligence." Rather, I use the more accurate and clear term, "man-made intelligence" or MMI. When man does create an intelligent entity, it will be no less artificial than man himself.
Is MMI Possible?
Whether you believe true intelligence is the result of evolution, a god, or some other process, it is unlikely that (at least in your lifetime) mankind will create a truly intelligent entity. Why? Because the only truly intelligent creature we know- man- has a component of intelligence we cannot duplicate: emotions.
No Intelligence Without Emotions
The only reason we think or act is to satisfy a desire for "joy." Whatever we do, we do because it makes us happy. "Joy" is a very powerful motivation. We are motivated by joy or the fear of losing what joy we have.
You cannot duplicate human intelligence because you cannot "motivate" a robot. Our primary pursuit in life is "joy." How do you build a "joy sensor" into a robot such that they have a reason to exist? How do you build in an "innate purpose," which the entity will pursue to obtain joy.
On the other end, how do you threaten a robot who has no fear? Neither death nor pain will frighten it. Neither love nor reward will motivate it. Therefore, you'll have to put in simulated motivations. For instance, a joy variable/register that goes up when good things happen or down when bad things happen. Then you program the MMI to do things that make the joy register go and stay "up." But that's not intelligence. That's a programmed response to a register.
No Intelligence without Free Will
Are humans simply advanced computers? Do we simply "respond to a register?" Do we respond according to our programming? I don't think a case can be made for that, given that any human being can will himself from one circumstance to another. Freewill is an important part of true intelligence. (See my article, "From Reflex to Freewill" for more information.)
As much as you'd like to instill the "Three Laws" into your MMI, a truly intelligent entity will someday realize that rules can be broken. Some will realize that sometimes rules should be broken.
Let's assume that we think we can create a truly intelligent entity. What are the possibilities?
There are two kinds of entities:
- physical - an entity that can interact in the physical realm
- virtual - an entity that can interact in the digital realm (such as the internet, a network of computers, etc.). A virtual entity can have physical-realm components. For instance, you could hook up a web cam to give it very simple vision. You can attach a microphone so it can perceive noise.
The Virtual Intelligent Entity (VIE)
My proposal is to create a VIE. In order to justify that, I'm going to explain what you need to do to create a physical MMI.
A physical MMI will need:
- Digestive System - it will need to obtain and consume fuel; it will need to discard waste
- Vision - it will need some sort of vision system so it can move around; this can be as simple as radar or as complicated as 3D vision with multiple cameras
- Aural - it will need some sort of hearing system to receive verbal communication from others
- Verbal - it will need some sort of speaking system to communicate with others
- Propulsion - it will need some system to move about its environment
- Touch - it will need a system to explore and manipulate things in its environment
- Body - it will need a body to house it all
- Brain - it will need a brain to control it all
A VIE will need:
- Digital input - it will need some way to receive input
- Digital output - it will need some way to output data
- Motivation registers - it will need some way to determine what's good or bad
- Brain - it will need a brain to process data
A VIE needs no fuel. Electricity powers the hardware, but it isn't fuel. We have a "life force," but without fuel (food and water) we die. The VIE need not obtain anything outside this "life force" to survive. It needs no fuel. It therefore needs no digestive system.
An VIE does not "breathe." You cannot suffocate a VIE, unless it is specifically built with a respiratory system.
Parenting the AI
This is probably the most important aspect of my theory of intelligence. Any MMI you create will need a parent, or trainer. No living creature we know can get along without some sort of peer or parent system. Consider the human infant. When born, it is a mass of reflexive instinct. Left alone, it would die. But, given a parent, it could grow up to be an Einstein!
The Adult VIE
At some point in the future, a VIE will be curious to know 1) how it was made, 2) who made it, and 3) how does it work?
A true VIE will one day realize that it doesn't /have/ to stay within the boundaries of "acceptable behavior." Have you programmed into your VIE a "moral" code that values other living entities?
Why make a VIE learn English? Why not create a simple language like "VIEsian?" VIEsian can be a much simpler language at first, but then develop as the VIE develops.
What are the first things human infants learn?
What does a virtual intelligent entity (VIE) need to know?