Lenhart Schubert es profesor de la Universidad de Rochester y reputado investigador en las áreas del lenguaje, representación del conocimiento, inferencia y planificación. El video que presentamos en esta entrada, perteneciente a una presentación en el centro Turing, el profesor Schubert explica los retos a los que se enfrenta la Inteligencia Artificial en cuánto al conocimiento, y cómo la resolución de estos problemas puede ayudar a lograr el sueño de Turing, una máquina capaz de una interacción y respuestas inteligentes, como las de un humano.
Como trabajo os dejo el que veáis el video al completo y que hagáis un pequeño resumen crítico que podéis poner como un comentario en este mismo post.
Como trabajo os dejo el que veáis el video al completo y que hagáis un pequeño resumen crítico que podéis poner como un comentario en este mismo post.
3 comentarios:
The Turing´s dream, a machine that is capable of beginning a conversation with a person withouth this one isn´t capable of noticed that, it´s for now, only a dream.
The researchers on artificial intelligence have achieved till now big advances in this field. Nevertheless, the obtained advances have been applications for very specific tasks like filtered of email, detection of frauds for credit cards, …etc. These applications lack the common sense or a deeper reasoning, central point of Turing's dream.
The concept of conversation implies big knowledge about the world in general, even the most simple conversation. This is due to the high number of existing things in the world, since they are structured, since they can be used or since they can exist, the own nature of the things, its behavior, its perceptions, … etc., and even different things that every person perceives, sees or deals of different form to other one. All that implies that if we were achieving an estimation of everything what a person knows and what can appear in a conversation, it would give us like proved a high number of things to know.
Therefore, the challenge which it faces the experts, not only implies the representation of the knowledge of the world, but using this knowledge to interpret the language, the processes of reasoning and the accumulation of the above mentioned knowledge. For the first one, already has seen before the challenge so important that it carries to try to represent all this knowledge. To interpret the language, is used the knowledge represented by means of the first order logic, a more precise representation of the language obtained from operators and logical formalisms using mathematical reasoning; the problem of this representation lies, how its sees in the video, in the inability to represent a more ordinary knowledge that requires a serial of extensions, that it gathers in the so-called FOL “language-like” extension. All this only for the part of the knowledge representation, the challenge is center in the representation of this language. The problem not only lies in the creation of a machine that contains all of this knowledge and the creation of all the necessary for communicate with it, but the representation of “languaje-like” that simulate the language uses by a human being. The process will be, more or less, to calculate the logical initial form and to represent it in this form, which must be not ambiguous and dependent on the context. Since it’s appreciated, the whole process depends always on the acquired knowledge. The great problem of this process takes root in that the human language is inherently ambiguous and the possible current solutions, as the representation by means of primitives, they do that they are slightly useful for the present time. The great objetive, therefore, is the interpretation of the natural language using the knowledge. In the part of the processes of reasoning, the most brief part of the video and probably the most important, appears like it's necessary to be the process of reasoning from the acquired knowledge, where major importance receives the not deductive reasoning. Finally, to complete the challenge, it's necessary to solve the problem of the accumulation of the whole necessary knowledge, as well as the neck of bottle that is generated on having tried to treat all this knowledge.
Since it's mentioned in the video, Turing's dream is attainable, is a difficult, but possible task, not for the present time.
La máquina de Turing: una máquina capaz de mantener una conversación con una persona sin que esta reconozca a la máquina y crea estar hablando con cualquier otro ser humano.
Este es el "Sueño de Turing", del cuál todavía distamos mucho a pesar de los avances obtenidos en la investigación dentro del campo de la IA.
El gran problema es la enorme cantidad de información que es capaz de almacenar y manejar el ser humano y que una máquina debería poder procesar. Las investigaciones actuales sin embargo se centran en resolver problemas concretos, cuando el problema radica en como representar todo el conocimiento y como usar este conocimiento para interpretar el lenguaje y para razonar correctamente.
Usando las matemáticas como herramienta, los investigadores se basan en los símbolos para poder representar el conocimiento en general, pero se requiere de un modelo que permita realizar esta representación. Un objetivo primario es la representación del lenguaje natural utilizando el conocimiento, lo que no es tarea fácil dada la enorme cantidad de información disponible.
En sueño de Turing se hace desear... pero seremos capaces de vivirlo por él.
Is possible a machine that can speak with a person, and a person doesn't know that the person who are talking with him is a machine. It is possible?
That is the Turing's dream, the first step to obtain this capacity on a machine, is that a machine be able to manage a big amount of data,and like a person also the machines have to be able to reasoning with the knowledge that they have stored. Nowadays all the advances, are aplied to other fields.
I want to see this moment,when we can speak with a machine and she don't reply with "I don't understand you, please repeat his choice". And then we have to do the Voight-Kampff test like in "Blade runner".
Publicar un comentario