User:Purple.Paanda/Progress in artificial intelligence

Article Draft
In his famous Turing test, Alan Turing picked language, the defining feature of human beings, for its basis. This test consists of a machine, in this case an artificial intelligence, tricking a human evaluator into believing that it is human. This is possible by having the evaluator engage in a natural language conversation with the machine and a human participant, of whom he does not know the distinction at first. The test can be considered as successful if after the conversations, the evaluator is still unable to tell which participant was human.

However, this test comes with a few flaws. For example, there could be many types of intelligence that would not pass the Turing Test. It is a very human-centric view of intelligence, which in other words, states that it assumes human intelligence is the only type of intelligence that it can measure. Also, it is not an objective test as it relies on the subjective judgment of the human who is evaluating. This induces biases and inconsistencies in the results. There could be different conditions depending on who the evaluator is, which makes it difficult to compare results across different tests.

Moreover, the Turing test is now considered too exploitable to be a meaningful benchmark.

More recently, the ACT test was created. It supersedes the Turing test. This test asks certain personal questions that an ordinary non-conscious machine would ideally not be able to answer. For example, some questions consist of “What is it like to be you at this moment?”,; “What was your experience seeing red for the first time?”, etc. An AI would pass the test if it could convincingly tell us about its consciousness and that we knew that it hadn't had any previous access to descriptions of consciousness.