Since the release of ChatGPT, it has won various praises and is praised by people from all walks of life. Bill Gates spoke highly of the significance of the emergence of ChatGPT, which will be no less than the birth of the Internet and personal computers. On the Internet, netizens rush to experience its intelligence: writing essays, writing college entrance examination essays, writing code, writing novels...
There are also rumors that ChatGPT has passed the Turing test. After searching on the Internet, it was mostly speculation from all walks of life, and there was no clear news.
AI is evolving rapidly, and its maturity is increasing. Gartner released the 2022 AI Technology Maturity Curve, which reflects the current state of AI development.
Of course, there are many ways to verify the level of artificial intelligence, such as the Turing test and the Winnograd test, etc.
First, the Turing test
The "Turing test" is one of the most famous tests in the field of artificial intelligence, first appearing in a paper published by Turing in 1950 called "Computational Machinery and Intelligence", which is a set of methods to determine whether a machine has artificial intelligence.
The Turing test is the original concept of artificial intelligence, which predates the word "artificial intelligence" itself, which was only proposed in 1956. Alan Turing, the inventor of the Turing test, is known as the father of computer science and the father of artificial intelligence.
The method of the Turing test is actually very simple, that is, to isolate the tester from the test subject (a person and a computer), and ask the test subject questions through some devices (such as a keyboard). After multiple rounds of testing, if more than 30% of the testers cannot determine whether the person being tested is a human or a machine, then the machine passes the test and is considered to have artificial intelligence.
Second, the Winnograd test
Similar to the Turing test, which aims to assess the intelligence level of machines, the University of Toronto has proposed the "Winnograd" test.
Winograd Schema Challenge (WSC), also known as the Winograd Schema Challenge. It is a machine intelligence test proposed by Hector Leveske, a computer scientist at the University of Toronto, in an attempt to improve the traditional Turing test. It is detected by asking the machine a specially designed multiple-choice question. These problems all contain a special structure called the Winograd Schema, named after Stanford computer scientist Terry Winograd.
In the test, the machine needs to indicate the antecedent of a pronoun in the question. In order to answer questions correctly, machines need to have the ability to reason with common sense.
The Turing test is to determine whether a machine can think and show intelligence that is indistinguishable from a human. It gives an actionable definition and provides a set of objective criteria for judging intelligence. However, in actual testing, the machine under test will deliberately respond with bluffing and confusing answers. Some of the programs participating in the test pretend to be crazy and stupid, some use rhetoric to interrupt the interlocutor's train of thought, etc., just to pass the test.
It was also in response to this phenomenon that the Winnograd test was proposed as an alternative to the Turing test. Compared with the Turing test, it pays more attention to the machine's ability to understand common sense reasoning and language subtleties. It can better detect whether the machine has intelligence in a deeper sense.
Of course, whether or not these tests are passed, the most important thing is to see the performance of artificial intelligence in practical applications.
Not to mention whether it passed the test or whether it had human intelligence. At this stage, the scientific and technological innovation and commercial application innovation brought by it will continue to penetrate our work and life, and the social impact will also be huge.
#新人小白求关注 #
#所见所得, it's all scientific #
#chatGTP人工智能 #
#头条新人 #