Friday, July 25, 2008

Popular on AI

Environment us nature, society, the universe - it is chaos or did something orderly? Can you describe or explain all the key number of formulas and algorithms? In fact, there is no necessary in qualitative assessment, which need to get rid of uncertainty. The very same uncertainty arises in the absence of the necessary number of data and relationships? If you are an experienced player in the chess situation on the board looks quite deterministic, that is predictable, because while he appreciates the many chess indicators, it all looks for Beginners uncertain. Even more, some of the indicators that you want to evaluate, and not even he suspects. But there is another side issue. For its turn to discuss literary classics, namely, the product of MA Bulgakov "Master and Margarita". The situation with Berlioz and Annushkoy, which shed oil. At the terminology AI in terms of higher forces both characters on the law can be regarded as agents who do not even know about the existence of each other, the more the relationship…

Yes, in fact, speaking of artificial intelligence, it is difficult to keep from having to go to the level of philosophical narration with a fantastic mystical classes. Although… Our readers love fiction, so let's look at an interesting example. Moreover, it is of direct relevance to practice.


During the period from 1961 to 1967 was published very popular fantastic saga of the three novels "Reason", "Foundation and Empire", "second base". Their author is practically our compatriot (he was born under the Smolensk) Isaac Azimov. The whole story begins with the fact that during the existence of a powerful empire appeared science such as psihoistoriya. Quoting that under this understood: "a branch of mathematics which deals with the reactions of human societies on a stable social and economic incentives… using precise mathematical data can be a certain way to influence human society". So, by using complex mathematical formulas that take into account the large amount of data, expressed in a number of factors, one of the scientists accurately predicted the fall of empires time as such - exactly five years.

And now is to think: Does it really? That is offensive to predict any developments in time. In principle, it is even possible. For example, in calculating the ideal model of development for something, and then begin to enter data with coefficients of influence. In principle, everything is overdone, although people familiar with games in the genre of economic policies, particularly the level of Pharao, Caesar, CivCity: Rome, Capitalism, etc., clearly understands what is at stake. For others cite another example, a medical. What can be predicted probability of occurrence of a disease and even predict the approximate time of the attack, if there was no question about infections? To do so, must obtain the largest possible number of necessary input data, which are produced as a result of analysis and research, plus take into account the aging of the organism, seasonal aggravation (for example, gipertoniki suffer in the heat), heredity, abuse, ways of life. As a result, real prediction. And the results can be and in reverse. What, indeed, modern medicine and works. If everything translated into mathematical language, then many of us around processes and events have inertia, so you can make some projections, using extrapolation, that is, if we are talking about a temporary stage, it is the spread of previously identified trends with respect to future periods . After all, often facing our ordinary person with statistical projections. In doing so, you need to take into account many different factors, sometimes in direct contradiction. For example, the growth of the working population is linked to improving the quality of life, that is good food, medicine, etc. But while in developed countries, fertility is low. And so on. There are such well-known subjects as management theory and statistics, which developed in the framework of their experience and achievements, in turn, transferred artificial intelligence systems at the expert level. And all this can be attributed to other quite different areas. And on the shoulders AI might well be tasked to ensure the viability of the same rights.

The usefulness and reward

The main task of agents decision-making often is to find the best option from a variety of alternatives. Let us turn to the same chess, which we have written quite a lot of material in the past, and let us assume that each of the figures has its price - for example, a pawn - 1, horse and elephant - 3, rook - 5, queen - 9, King - 1000 . In fact, the program works for reward, as if asking: "What do I get for it?" That is developers realized bonus system, expressed in so-called utility function - the term used in AI. There need to understand that this is not what an agent can bring benefit to their actions or state of choice for others. He - rational egoist - in fact, none other than himself, does not see and always wondered what benefit this or that decision could bring him the most. In the biological nature all very similar. For example, in the human brain has a pleasure center, and so on. Thus, a motivation to find the most rational result. Why have we turned to chess? There, as the agents AI, the actions are in the mode of uncertainty. In fact, expected to feature a chess program and the utility function with agents identical. Arriving at the instance, branches, including the taking, the program expects options for the largest possible number of points. And this is the most primitive and trivial form. Recall that Deep Blue six monitors thousands of chess-specific indicators. If this computer algorithms in the evaluation function only take into account takeover, the computer would never have been able to win rights.

The interim phase

Before turning to the specific tasks and presentation of practical bases, I suggest you independently perform a number of assignments. Basically this answers to the questions:

. Try to independently withdraw several definitions of artificial intelligence.
. What agents or their community can be interconnected assigned to the intelligence systems?
. Try to imagine the situation reversed with Turing test, namely: how and on what grounds the computer can "understand", that people are communicating with him?
. Can I take it that artificial intelligence above human if he had received the highest score in tests of IQ?
. Can I take the affirmative by the fact that semantics (a full understanding of the meaning of the words and phenomena) in humans is directly related to identity, that is, comes from a personal "I"?
. And whether it is possible to create such a framework II, focusing only on the utility function and implement a system of rewards?
. Imagine what kind of object AI is on a random cell chessboard. Think algorithm, which he will be able to quickly find the coordinates of his location, that is to specify cell.
. Children's play "rock, scissors, paper" refers to a deterministic or nondeterministic type?
. In the popular game Starcraft professionals identify three key strategies of development armies, one of which provides a quick victory in the initial stage of the game, but almost always lose later, the second - on the contrary, and so on. Can I hold in this case a parallel with "rock, scissors, paper" in the strategic sense?
. Society Society of Automotive Engineers (SAE) is the developer of architecture JAUS (Joint Architecture for Unmanned Systems), implying a single management protocol Unmanned robotics. That is thanks to him, you can manage robotankami, robosamoletami, robosoldatami, roborazvedchikami, autonomous vehicles, using a unified command system. What requirements must present to develop AI to completely transfer the administration on its basis?
. Among the many different and not very plausible theories can be found and such: the main inhabitants of our planet - a bacterium, and all life is a container for them. Oprovergnite and prove this theory in terms of the vision of community of bacteria as intellectual substance. And whether the submission in principle?

No comments:

Post a Comment