Turing's test was developed by the scientist Alan Turing and includes an experiment in which the participant cooperates simultaneously with a computer and a living person. Based on the answers received to their questions, the participant must determine who they talk to: man or machine. If the unit could not distinguish between them, it was decided that the machine successfully “underwent” the test.
However, this test, once considered innovative, now has its limitations. He focuses primarily on imitating human reactions, and not on true human reasoning. Many models of artificial intelligence are distinguished by imitating conversational styles, but often there is no deep mental abilities. This does not require that artificial intelligence has self -awareness or understand your own reasoning. Even Turing admitted that this test cannot really determine if the machines could think; It's more about imitating than getting to know.
Earlier, we studied the GPT-4 problem, which undergoes Turing test and the results obtained from such an experiment. You can read the article here.
To deal with the mentioned restrictions on Turing's test, Philip N. Johnson-Laird from Princeton University and Marco Ragni from Chemnitz University of Technology have developed an alternative to a known test. They suggest a shift of focusing on the basis of whether the machine can imitate human reactions to a more fundamental question: “Does AI reason in the same way as people?”
Their Published article It presents a new assessment system, which aims to determine whether artificial intelligence really reason like a human. These frames consist of three key steps.
1. Test the program in a series of psychological reasoning experiments.
The first step is to conduct a series of psychological experiments between AI models that are distinguished by human thinking from standard logical processes. These experiments delve into various aspects of reasoning, studying nuances that differ from standard logical frames.
If the judgments of the machine differ from human judgments, we answered the previous question. Computer differently from people. However, if his judgments are much consistent with human reasoning, we go to the other step.
2. Test understanding the program of your own reasoning process.
This step aims to assess AI understanding of your own reasoning processes, a critical aspect of human cognition. Ideally, if the machine was able to analyze its reasoning and explanations of its decision, reminiscent of self -analysis similar to human behavior.
If the program passes this test, the third step is analytical.
3. Check the source code of the program.
The last step includes testing the source code of the program. If it contains the same basic components that are known to model human activities, including an intuitive system of quick deductions, a system that is prudent for more thoughtful reasoning and a system of interpretation of terms based on context and universal knowledge, this evidence is crucial. If the source code of the program reflects these rules, it is believed that you think like a human.
Considering AI as a participant in cognitive experiments, this innovative approach means a change in the paradigm in the assessment of artificial intelligence. Subject to the computer code for analysis, scientists propose a re -assessment of AI assessment standards. Because the world still strives for more sophisticated artificial intelligence, this new concept can be a significant step forward in our understanding of what the machine thinks.
