What do you think?
Explore Procon's discussion
Artificial general information (AGI), or powerful AI, that is, artificial intelligence intended to replicate human intellectual abilities, is controversial and out of reach. The difficulty of expanding AI's modest achievements cannot be overstated.
However, this lack of progress is merely a testament to the difficulty of AGI, not an impossible possibility. Let's look at the very idea of Agi. Do you think a computer probably? Theoretical Linguist Noam Chomsky suggests that it is pointless to discuss this question. Because it is essentially an arbitrary decision. think Include machines. Chomsky has no de facto doubt as to whether such a decision is right or wrong. Because I have no doubt about our decision to say that the plane would fly is right or that the boat swim is wrong. However, this seems to oversimplify the problem. The important question is that it is appropriate to say that the computer is thinking, and if so, what are the conditions that the computer must meet to explain it like that?
Some authors provide Turing test as a definition of intelligence. However, mathematician and logician Alan Turing pointed out that if it cannot imitate a human well, it may nevertheless fail the test. For example, ChatGpt often calls status as a large language model, so it's unlikely to pass the Turing test. If an intelligent entity is able to fail the test, the test cannot act as a definition of intelligence. As information theorist Claude Shannon and AI pioneer John McCarthy pointed out in 1956, passing the test is even doubtful if a computer actually shows that it is intelligent. Shannon and McCarthy argued that, in principle, it is possible to design a machine that includes a complete set of canned responses between all questions and between all questions. Like Parry, this machine generates answers to interviewer questions by looking up appropriate answers in a huge table. This objection appears to indicate that, in principle, systems with no intelligence at all could pass the Turing test.
In fact, AI does not have an actual definition of intelligence it offers, even in subhuman cases. While mice are intelligent, what should artificial intelligence achieve before researchers claim that they have reached the level of success for rats? If there is no reasonably accurate standard for when an artificial system is counted as intelligent, there is no objective way to communicate whether an AI research program has succeeded or failed. One of the consequences of AI's failure to create satisfactory standards for intelligence is that it is a program that allows researchers to hold conversations like GPT, for example, or beat world chess champions like Deep Blue.that's right It's not intelligence! ” Marvin Minsky's response to the problems defining intelligence is that, like Turing before him, intelligence is simply our name in the mental process of problem solving that we still don't understand. Minsky likens intelligence to the concept of “unexplored African regions.” It disappears as soon as it is discovered.
