I wrote a post for LJ, which led to some interesting discussions, so I figured I may translate it to English and post here to see what you guys think about that.
We, humans, certainly believe that we are intelligent. It is a phylosophical question whether we are, but we undoubtedly posess some of the attributes of intellligence: language, use of tools, cultural and social institutes and so on... I can argue that these attributes and institutes are result of evolution and that micro-societies (tribes?) that did not posess them were more likely to die out.
Now if we look at DBs, we can see that (at least in theory) similar institutes can be developed and evolved. The sims are simple right now, but I have already seen disappearance of cannibalism in my sim. Given enough complexity we can imagine situations in which almost any our social institute has an advantage.
Now let's get back to intelligence. We wouldn't call bots intelligent simply because they are not cannibalistic. But what if they have complex language capable of describing abstract concepts? What if they build houses? Have kindergardens? Write books? Dance? Pray? Make computers and run their own artificial life sims? Would you at some point call these bots intelligent? Don't answer yet, it gets a bit more complex.
DBs (and any other art life sime) is a deterministic system. Give it the same seed for randomizer and the sim will repeat itself perfectly (there are more complex ways to introduce randomness, but let's not go there yet). Now imagine that you were talking to a bot about the meaning of life and had a perfectly reasonble conversation. However, if you start the sim over and ask the bot the same questions at exactly same point in sim, then you will get exactly same answers. It becomes obvious that you were not talking to an intelligent being, you were talking to a very complex program. Give it inupt "A" and it will always return "B". That is not intelligence.
But why? What is different? Well, we intuitively feel that we, humans, on any given input "A" can give a vast number of outputs. The person asking the question has no way to know in advance, which answer we give until we do. So is the answer - unpredictability? But unpredictability by whom? Bots can't predict each other's answers, they don't have enough information on how DBs works. But they are predictable to us. So, we need to make them unpredictable to us, humans. How? Well, we can hook up a Geiger counter to computer and seed a truly random number into the sim at each and every cycle.
Ingenious? No, it is not. Think about it - if before you were talking to a very complicated program, now you are talking to a very complicated generator of random answers. Right? 10 electorns hit geiger counter, the answer is "B", 20 electrons - "D".
This is not intelligence.
In order to consider the oponent intelligent we must assume that it is capable of giving many different answers, but it does not choose the answer randomly. This is described by the concept of "free will". The problem, however, is that "free will" is a phylosophical concept and does not have a measurable manifistation in our world. We can only state that we feel free will in ourselves, but we have no clue about others.
As a reuslt, our definition of intelligence relies not on some scientific observation, but on a belief in existence of free will in other people. We believe that they have free will, they believe that we have free will. Mutual agreement to stabilize our psyche. The most basic and, probably, the most ancient sociological institute in human history.
So, where does it leave us? Still in the same place - intelligence is undefinable by scientific methods and is a matter of opinion only.
Ironic, isn't it?