Artificial Intelligence
Farid Fleifel Tapia
faflta@inf.upv.es
Index:
To define the Artificial Intelligence is a task actually complicated. From that it is acuñara this term in the Conference of Darthmouth of 1956, it is they have proposed many definitions different that in no case they have achieved the aceptación without reservas of all the community investigadora..
This puede it be must to the insuficiente formalización of the matter, that prevents a asentamiento firme of the bases of this science. So much is thus that in ocasiones the expert neither even coincide in the appropriateness of certain problems to the tratamiento through technical of intelligence Artificial, is to say, some times neither even it is has the safety of that a problem given it has to be studied within of this disciplina.
For luck this it is changing, and each time more, the technical of AI, furthermore of to acquire a base formal acorde with the needs of the matter, they are letting of be "curiosidades académicas with much future but few application in the present", for to happen to be one of the motor that impulsan the industry of the computation, with investments crecientes year after year.
The possibilities from then they are astonishing, but not there is that let volar the imagination more of it necessary. The "machines pensantes" puede that arrive to to exist some time, but not they are to the return of the esquina neither much less. The AI is a science new that it must andar much road still. Of it that we can be seguros is of that the first steps they are being promising.
We could to say without dread to errar that the AI is the branch of the science of the computation that center its efforts in the attainment of systems intelligent, but this definition is so vaga that, not aclara actually neither which they are the objective of this science neither which the methods that it is they should to apply for to reach those objective. In these two points is in the that hallamos the controversia.
[ Back to Index ]
In first place, ¿What is it that it is wants to say when it is assert that the objective of the AI is the attainment of systems or "entities" intelligent?. ¿For that a entity it will be considerada intelligent enough with that it is behave inteligentemente, or furthermore it must to reason of form intelligent? Puede that the difference between the two possibilities seem sutil to first sight, but has more importance of the that seem.
For example, a program of chess acts inteligentemente, but in essence not puede it be said that razone, already that it only that makes is to explore a great number of possible plays -million in each shift-, applying a function of evaluation that says it good or wrong that is each a of those plays, having in account for this the number of pieces of each player in the checkerboard and the value ponderado of the same. To to depart of this information, in each shift it is chooses the better of the plays obtained. In contraposición, the Large Maestros of chess only analyze some few decenas of plays possible in each movement, and without embargo until the moment the machines of chess, with its great power of calculation it they have had complicated for vencerles. The difference rest in the form of to analyze the plays for part of the Maestro human and of the machine. The machine is pura force bruta. Analyze all or the greater part of the possibilities, without to have in account in first instance the viability of the same in the context of the game. The Great Maestro human not is so good in the exploration intensiva of plays, but accomplish much better the task of to eliminate of its consideration the lines of game little promising, a task much more complicated and that exige of a comprensión and vision of the game that the machines not they have obtained emular still. Arrived to this point we can to admit that a program of chess of last generation acts of form intelligent, already that its game is of a level more high that the of the majority of the persons, but ¿Is actually intelligent?
As we will see more forward the philosophers of the AI have several answers for these incógnitas. We of moment us pondremos a goal relatively humilde and we will say that a technical of AI is that that it is use for to achieve that a program it is behave of form intelligent, without to have in account the "form of reasoning" underlying to the methods that it is apliquen for to achieve that behavior.
Here returns to to emerge a dilemma. According to this definition, to first sight parecería that almost any problem resoluble for computer would encuadrarse in the field of the intelligence artificial. For example, we suppose that us aplicamos in to solve the problem of the division of two numbers racionales. A simple calculator of bolsillo solves this problem without greater complicaciones, and it makes so well as a be human, giving a response intelligent, or more well congruente, to our petitions. Without embargo, nobody diría that the calculadoras apply technical of AI for to accomplish the divisiones, as seem to suggest in principle our definition, of manner that we will will have that cambiarla or to the less retocarla. The solution definitive llegará after the definition of a par of terms fundamental, task that abordamos to continuation.
The problem of the division puede resolverse applying a kind of receta with some steps well defined, that always operate and that all the world learn in the first years of colegio. In the science of the computation to these recetas that serve for to solve problems it is you flame algorithms, term acuñado in honor of the mathematical árabe To the-Kwarizmi, that between other things compiló a ready of these recetas, applicable to different problems algebraicos.
Since well, when someone apply one of these algorithms, in essence not it is acting inteligentemente. Anyone puede to find the response correct, already that enough with to follow the rules to the foot of the letter. For to understand this enough with remontarnos to the first years of school, when aprendimos to to divide. To the majority us occurred in a principle that aprendimos the receta and descubrimos surprised that yet without to know exactly it that estábamos making, if aplicábamos the rules to the foot of the letter obteníamos the solution correct. Not there is that be especially intelligent for to know to divide. Only there is that to have the capacity of memorizar some how much rules and aplicarlas always of the same form, thing that a ordering is capable of to make much better that the persons.
The problems as the of the division, that have a solution algorítmica known and effective, not they are the objective of the AI. Not import it complicated that it will be the problem neither how many rules there is that memorizar and to apply. If the problem puede resolverse of form exact for middle of a algorithm, consideraremos that not there is intelligence implicada in the resolution of the same.
[ Back to Index ]
The problems actually complicated to the that it is face the be human they are those for the which not exist algorithm known, well because not it is there is overdraft yet, or because definitivamente not exist -there is problems for the that it is knows that not exist a algorithm that the resuelva. Hablaremos of they more forward-.
We put as example the problem of the calculation of the primitiva or "integral" of a function, This is a problem for the that not exist algorithm. Exist some rules that they can seguirse for to obtain the primitiva of certain functions with some determined characteristic, but not exist a algorithm genérico that us diga what steps there is that to follow for, departing of a function anyone, to find its primitiva. The mathematical it is bases in certain rules that generally operate, or that in many cases serve for to reduce the complexity of the problem, converting functions complicated in other more simple that yes they are capable of integrar. Without embargo the application of these rules not garantiza the obtainment of a solution in all the cases.
The rules of this type, that try of orientarnos toward the solution in problems not algorítmicos they are calls heurísticas. These rules they are so concretas as the dadas in the algorithms. The difference reside in the fact of that nothing us garantiza that the application of a heuristic us carry a step more near of the solution of a problem. Furthermore, it is gives the situation of that in many ocasiones exist a great quantity of heurísticas that they are applicable to a same state of the problem, with it which the election of the more adapted for each case puede be a factor determinant in the obtainment of the solution.
These they are the problems actually complex and that need of real intelligence and originalidad for its resolution. Not there is algorithms known that us permit to discover theorems mathematical, diagnosticar diseases, componer music with armonía and gracia or to understand a phrase written in a language natural as the español or the english, though yes there is certain heurísticas that the people apply for to accomplish these tasks, even to times without it be given account. These they are things that the persons make without to know actually how the make neither what steps continue in all the cases, for much introspección to the that it is sometan.
And these they are the tasks of the that it is occupy the AI.
Of manner that to the sight of all these new data, redefiniremos our concept of AI of the following form:
We will say that the AI is the branch of the science of the computation that study the resolution of problems not algorítmicos through the use of any technical of computation available, without to have in account the form of reasoning underlying to the methods that it is apliquen for to achieve that resolution.
It is observará that, to difference of the definition previous, in this not it is mention explicitly the behavior intelligent of the technical of AI. This is thus because it is asume that any process computacional capable of to solve instances of a problem not algorítmico for force it must to give the appearance of a behavior intelligent.
This definition puede to seem a so much críptica for a not iniciado, and furthermore not has garantías of be accepted for all the community of investigating of AI. For it so much, and for complementarla, we go to to present, without comentarlas, some definitions something less formal, dadas for different investigating in AI, that they are compatibles in great measure with the given for we, but considering other points of sight.
The AI is:
The art of to create machines with capacity of to accomplish functions that accomplished for persons require of intelligence. (Kurzweil, 1990).
The study of how to achieve that the computers accomplish tasks that, for the moment, the human make better. (Rich, Knight, 1991).
The branch of the science of the computation that it is occupy of the automatización of the conduct intelligent (Luger and Stubblefield, 1993).
A field of study that it is enfoca to the explanation and emulation of the conduct intelligent in function of processes computacionales (Schalkoff, 1990).
A time definida the AI as branch of the science of the computation, deberíamos to have in account the methods that it is will follow for to achieve the objective that us we have proposed. The sustrato of the AI it form the machines of cómputo and process of data digitales, already that these machines they are the dispositivos of processing of information more complex and potent that there is developed the be human, and the activity intellectual is in essence a process of manipulation and analysis of information.
It which not wants to say that the utilization of other paradigms of processing it be descartada. Is more, having in account the tremendous yield that our brain, slow, analogous and asíncrono, obtain in these tasks of processing of information, perhaps deberíamos to consider other possibilities. The nets neuronales, paradigm of computation very in boga lately, exploran this road, trying of remedar the operation of the brain human until certain level, and consiguiendo spectacular results that auguran a future i fill of possibilities.
[ Back to Index ]