What do you think?
Explore the Procon debate
General artificial intelligence (AG), or a strong AI-that is to say the artificial intelligence which aims to duplicate man intellectual Capacities – controversial and out -of -reach remain. The difficulty in increasing the modest achievements of AI cannot be overestimated.
However, this lack of progress can simply be a testimony to the difficulty of AG, and not its impossibility. Work us towards the very idea of act. Can a computer think? The theoretical linguist Noam Chomsky suggests that debating this question is useless, because it is an essentially arbitrary decision to extend the common use of the word think To include machines. There is, says Chomsky, no factual question as to whether such a decision is good or wrong – just as there is no doubt that our decision to say that planes are stealing is right, or that our decision not to say that ships swim is bad. However, this seems to simplify the questions. The important question is as follows: could it be appropriate to say that computers think and, in the affirmative, what conditions should a computer satisfy to be described?
Some authors offer the Turing test as a definition of intelligence. However, the mathematician and logician Alan Turing He himself pointed out that a computer that was to be described as intelligent could nevertheless fail his test if he was unable to successfully imitate a human being. For example, pussy often invoke its status of Great language model And it would therefore be unlikely to pass the Turing test. If an intelligent entity can fail in the test, the test cannot work as a definition of intelligence. It is even questionable if passing the test would actually show that a computer is intelligent, as an information theorist Claude Shannon And the pioneer of AI John McCarthy underlined in 1956. Shannon and McCarthy argued that in principle, it is possible to design a machine Containing a full set of answers keeps in all the questions that a questioner could possibly ask for the fixed period of the test. Like Parry, this machine would produce answers to the questions from the interviewer by looking for the appropriate answers in a giant table. This objection seems to show that, in principle, a system without intelligence could pass the Turing test.
In fact, AI has no real definition of the intelligence to offer, not even in the subhuman case. Rats Are intelligent, but that exactly an artificial intelligence should be achieved before researchers can claim that it has reached the level of success of rats? In the absence of a reasonably precise criteria Because when an artificial system also counts intelligent, there is no objective way to say if a research program on AI has succeeded or failed. A result of the failure of the AI to produce a satisfactory criterion of intelligence is that, each time the researchers achieve one of the objectives of the AI - for example, a program that can organize a conversation like GPT or beat the world chess Champion like Deep Blue – Critics is able to say: “It is No intelligence! “” Marvin MinskyThe answer to the problem of the definition of intelligence is to maintain – as Turing before him – that intelligence is simply our name for any mental process of problem solving that we do not do Again to understand. Minsky compares intelligence to the concept of “unexplored regions of Africa”: it disappears as soon as we discover it.