Fink About It

This is an informative essay about artificial intelligence. Firstly this essay will argue that artificial intelligence is nothing to be afraid about, and secondly that its development is driven by computer games. In the following I am going to explain how artificial intelligence works and how intelligence is created. To do this I will be looking at some representative movies and games to show how artificial intelligence is depicted. Finally I will investigate its restrictions and misbehaviors with the help of the “Tale-Spin-Program” which was developed in 1976 to invent animal fables.

Are you afraid of Artificial Intelligence?

Whenever people hear the term “artificial intelligence” mainly one thing comes to their minds: artificial intelligence means clever robots that think and act independently. Since there were movies like “the Matrix” and not to forget “I, Robot” this independence is not always a good thing. Things can get out of control and machines might conquer the world. The machines in those movies often conclude that humans are either dispensable for they destroy nature and themselves anyway (see “I, Robot”) or only useful to them in terms of energy (see “The Matrix”, where machines use humans as biomechanical batteries). Very scary, isn’t it?

However, there are other perspectives on artificial intelligence, mostly in animated sci-fi movies like “Appleseed”. So called “Bioroids” that look like humans help to stabilize society. Steven Spielberg’s “A.I.” will probably be the most prominent Hollywood example of artificial intelligence in a social context. The son of Henry and Monica Swinton is in a coma with very little chance of surviving. They agree to adopt a child robot prototype to fill the gap. Things get complicated when their real son survives by chance. So not all machines with sophisticated artificial intelligence have to be evil necessarily.

.¹

These two views contrast each other in the sense that artificial intelligence can be positive or negative. We must not forget, though, that this level of evolution is still up in the air. So let’s look at the definition of contemporary artificial intelligence.

What is Artificial Intelligence, exactly?

For a definition, let’s take the Merriam Webster Dictionary², artificial intelligence is defined in the following way:

1: a branch of computer science dealing with the simulation of intelligent behavior in computers

2: the capability of a machine to imitate intelligent human behavior

This imitation is done in three steps: Taking in information, processing it and finally doing something intelligent with it. To give you an example, I will explain this using the most famous search engine “Google”. First, you type in your keywords or a whole sentence. Then the words are processed by a clever algorithm that works its way through the indexed websites. According to your input, the location of the computer you are searching from, your language and whether you have enabled safesearch³, search results occur within seconds. Google’s search algorithm is believed to be the most effective one and therefore Google counts as the most intelligent search engine.

.⁴

Regarding the first part of my essay, there is one more point to make about artificial intelligence. Artificial intelligence is divided into weak and strong artificial intelligence. Machines conquering the world would have to have a strong artificial intelligence, which implies they have emotions and their own consciousness. We are still on the level of weak artificial intelligence that can only simulate a consciousness. Even with all the cameras and sensors on it, Sony’s Robot Dog⁵ remains a simulated, often misbehaving⁶ artificial intelligence. In this case there is no point in talking about strong artificial intelligence. The points made in this essay all refer to weak artificial intelligence.

How Artificial Intelligence works

Whereas the Google example (above) seems to be a very static one, the most dynamic artificial intelligence is to be found in computer games, where we compete with the artificial intelligence of the game. The question then becomes: Who is smarter, man or machine? Chess masters are still struggling with the chess program “Fritz”. In 1997 world champion Garry Kasparow lost against the IBM computer “Deep Blue”. He accused IBM of cheating. In 2006 world champion Wladimir Kramnik played against the chess computer “Deep Fritz”. However, he was equipped with the program beforehand and could prepare. Nonetheless, he lost 4:2. These two events shocked and stunned the chess world equally. The newspaper “Zeit” (12/06)⁷ even wanted to point out that mankind is not finished yet.

.⁸

The question now becomes how this man-(b)eating, simulated intelligence is realized. And the answer is very simple. It is programmed well. Programmers define how the computer shall behave for all possible moves. If the human chess player makes a move A, the computer takes in the information, processes it by looking into the artificial intelligence section and finding possible reactions to move A, and finally it responds as the programmer defined it to act on move A. Good artificial intelligence consists of a whole set of answers to A and any other move of the human chess player. The result – a computer beating a world champion – is what we think of as clever. But basically it’s only a set of the right variables, and the question arises if the human chess player could respond as cleverly as the machine when having the same access to these variables – not taking into account the processing time, of course. On the other hand, if there is no response defined for move A, the artificial intelligence will act badly or – in some cases – not at all.

To make this whole affair more observable, let’s look a typical artificial intelligence routine telling an ant if it is to attack another bug or not:

public override void see (ant bug)
{
if (numberofbugsinsight < 5)
{
attack(bug);
} else {
turnaround();
gostraighton(100)
}
}

So what does this routine do? Basically, it says that if there are 5 or more bugs in sight the ant is supposed to walk away 100 units⁹. In other words, we can also say that the ant is to escape, otherwise to attack.

Artificial Intelligence in Games

Experiencing the behavior of the ant fleeing from an army of bugs without looking at the code that is behind this, we can say that the ant is clever and does well to not attack the others¹⁰. Once again the success of good artificial intelligence is then to have all the necessary variables in it. Programmers have obviously made an effort to come up with all possible variables for a chess game, since it can beat complexly thinking world champions.

As long as it is as simple as in my example above, artificial intelligence routines can do a lot. Chess has 64 squares and 6 different types of figures that can be moved in a particular way. Therefore there is a finite set of variables to define in the artificial intelligence. Now think of a strategy game like “Anno 1701″ which consists of 1.000 x 1.000 squares. You can move ships and people on them, or you can build hundreds of different houses on each set of squares. According to “Anno 1701″ programmer Thomas Stein, “there is no computer in the world which can calculate the sheer infinite game situations that emerge from it.”

In another genre, namely shooters, it is important for the artificial intelligence to be a real challenge. They have to react to what the (human) player does. In the 90s enemies were just popping up approaching the player without caring what happened to them. In the last 4 years the opponents in shooters have become cleverer, using the things around them for cover, ducking behind a container and trying to surround the player. F.E.A.R. (2005) is the most famous example of having artificial intelligence controlled team manoeuvers. Having an artificial intelligence that smart, you can play the same game over and over, never having the same shootouts as before.

To enhance the complexity of artificial intelligence a lot of companies have specialized in developing artificial intelligence. One of them is Xaitment¹¹. This company develops artificial intelligence products that can be used by programmers for implementing smart artificial intelligence. These editors, which are split up into different artificial intelligence sections like think (autonomous decisions of NPCs¹²), know (world or individual NPC knowledge) and move (movement behaviors of NPCs), which – again – emphasizes the complexity of the matter.

I have to admit that this subsection was a bit one-sided. Nevertheless it is the artificial intelligence used in games which contributes most to the development of a clever artificial intelligence, since it is not restricted in terms of material or laws of nature. While AIBO is naturally limited by its size, its material and its pneumatic movement controllers, people and animals in games can fly, shoot, think, know without batteries or other boundaries, limited only by the amount of variables that can be calculated and the commands that can be executed by the computer simultaneously. Of course, there is also a downside in artificial intelligence and programmers have to use tricks like scripting a scene or drawing invisible lines for the enemies to not push at an open door, and so on.

Artificial Intelligence acting stupid

It is a fact that all that was said in the previous paragraph is simply the right set of variables for the artificial intelligence in a game. It can’t be denied that due to missing or wrong variables – and let’s not forget that this is the only way of simulating a clever artificial intelligence – the things artificial intelligence can do are very limited. Either it does not respond or, in the worst case, it responds the wrong way round. To illustrate this, I want to look at the “Tale-Spin-Program” which was developed in 1976 to invent animal fables.

The “Tale-Spin-Program” was fed with persons, plots and relations between them, so that it could come up with a story. It didn’t work out very well in the beginning, when the computer came up with the following:

One day, Joe the bear was hungry. He asked Irving the bird where he can find some honey and Irving told him that there was a beehive in the old oak. Joe got angry and clobbered him.¹³

The problem was that Irving had a rule saying if honey, then beehive whereas Joe didn’t have a rule saying if beehive, then honey. The program was corrected, but in the next step Joe went and ate the beehive. Later on Joe asks Irving where he can find some honey, but Irving doesn’t want to tell. Now Joe has a rule that tells him if you want to get something from someone you can either trade or trick.

Joe promises Irving a worm for telling him where he can find honey. Irving agrees and Joe searches for a worm, but he can’t find one. So he goes back to Irving and asks him where he can find worms. Irving doesn’t tell him. Joe, because of his rule, offers Irving a worm for the information where to find worms. Irving agrees and Joe searches for a worm, but he can’t find one. So he goes back to Irving asking him where he can find worms.¹⁴

The program ended up in this loop, so another set of rules had to be implied. When you cannot accomplish your goal, you may not try again, but focus on something else – otherwise you will get sick and When you see food, you have to desire it. The result was the following:

Henry the raven is sitting in a tree with a piece of cheese in his pecker. Bill the fox wants to have it. So he tricks Henry by asking him to sing. When the cheese drops to the ground, Bill the fox gets sick.¹⁵

So what happened? Bill saw the cheese in Henry’s pecker at first. When it dropped he saw it for the second time, still wanting it. The rule says that he may not have an agenda twice, so he becomes ill. The story continues as follows:

Henry sees and wants the cheese as well. To get it he needs to get it from its owner. The owner is he himself. One of his rules tells him that no one can outsmart himself. So he asks himself for it. But he doesn’t want to give it to himself. So Henry offers himself a worm for it, searches for it, but doesn’t find one. He turns to himself and asks himself where he can find worms. But he doesn’t know and replies that he can’t tell. To do this he decides to offer himself a worm, if he tells himself where he can find worms…

Conclusion

This essay has shown how artificial intelligence is derived. After all, artificial intelligence is something that is man-made and restricted by the variables that were or were not defined. I showed that movies presuppose a strong artificial intelligence that has its own awareness. Artificial intelligence, however, is an ongoing process and it is not yet anything to be scared about, for it can only simulate finite actions that were hard-coded in the first place. The last part of the essay has illustrated the complexity and the restrictions that occur with artificial intelligence and its programming. All things considered, our best bet on improving it is the ongoing artificial intelligence development in computer games.