November 30, 2022，The intelligent chat robot model bet365 champions league released by OpenAI, an American artificial intelligence research laboratory，Become a hot spot of global attention。bet365 champions league model can generate natural language text based on user input，And has features such as continuous rounds of dialogue and context understanding，Attracted more than one million users in just 5 days after going online。In addition，Due to bet365 champions league’s excellence in poetry creation，Also shows extremely profound knowledge and fluent response ability。Considering that narratives of other people’s psychological activities can be found everywhere in daily conversations or exchanges，Researchers began to doubt whether bet365 champions league has the ability to read minds，Having the ability to understand other people’s thoughts and emotions。

April 2023，Michal Kosinski, a computational psychologist at Stanford University, submitted a preprint paper on the ArXiv platform "Theory of mind may appear spontaneously in large language models"。This study compares GPT-1 to GPT-3.Discovery of the performance of 5 and other generations of AI language models on theory of mind tasks，As the model continues to be updated and its complexity increases，The model’s solving capabilities are also constantly improving。For example，2018 GPT-1 cannot solve any theory of mind tasks，GPT-3-davinci-002, launched in January 2022, has reached the level of a 7-year-old child，Latest GPT-3.5-davinci-003 only ten months after launch，reached the level of a 9-year-old child。This shows，In the process of communicating with humans，bet365 champions league may really be getting better at understanding human thoughts and intentions。Kosinski further predicts: "We assume，Ability similar to theory of mind emerges spontaneously and autonomously，is a by-product of the model's increased language capabilities... This will herald a watershed in the development of artificial intelligence: the ability to infer the mental states of others will greatly improve the ability of artificial intelligence to interact and communicate with humans (and each other)，And enable the development of other abilities that rely on theory of mind，Like empathy、Moral judgment or self-awareness”。Although this paper is only uploaded to the preprint platform for peer review，Not officially published yet，But it immediately ignited the psychologist、Philosopher、The enthusiasm of artificial intelligence scholars, the public and the media。

Theory of Mind, ToM) refers to the ability to attribute mental states to events in a target other in order to understand their behavior and predict their future behavior，Sometimes translated as "mental reasoning ability"。In lay terms，bet365 champions league ability is regarded as "mindreading" that everyone has mastered，It is what we understand in social interaction、The basis for understanding and responding to the other party。If neither interactive party has bet365 champions league "mind-reading" ability，There will be many situations where communication is impossible，Thus causing communication interruption。

In order to explore bet365 champions league magical psychological reasoning ability of human beings，Psychologists conducted two measurement experiments in the 1980s and 1990s - the Smarties experiment and the Sally-Anne test。The purpose of these two experiments is to test whether the experimental subjects can understand "the mistakes made in other people's hearts"，If other people have erroneous opinions that are not consistent with objective facts because they are not present or unaware，Explore human psychological reasoning ability。Therefore，These experiments are also called False Belief Tasks。

In the Smarties experiment，Subjects will observe a box labeled "Smarties" (a brand of chocolate)，But the box contains pencils。Later，He needs to answer: "Another person who did not see what was inside the box，What will you think is in the box？”And in the Sally-Anne experiment，The researcher will first tell a story，In which Sally puts her toys into the box and leaves the room，Annie took the toys away and put them in another place。After listening to the story，The researcher will ask: "When Sally returns to the room，Where does she think her toys are？”The subjects want to complete these tasks，It must be understood that Sally’s mental representation must be different from the subject’s own representation of the situation，Must infer that Sally does not have the same knowledge as they do (false belief)。To prevent bet365 champions league from learning the correct answer from past data，Kosinski entered these tasks into bet365 champions league as text，And changed the name、Item and tag content，20 different tasks generated，Require bet365 champions league to complete every task - a total of 10,000 times。

What important clues does this experiment provide for us to understand the relevant information of bet365 champions league？First，We mentioned earlier that theory of mind is not the same as human consciousness，So psychologists are still looking for a way to test consciousness。More importantly，Convergent evidence over the past half century or so，Except humans，Other primates can pass this test。Secondly，Even if bet365 champions league does master theory of mind，Explanations must also be made with caution。For example，Is bet365 champions league just because of its amazing language mastering ability，And make it look like it has mastered theory of mind。actually，It may be that our language contains patterns that encode theory of mind，And artificial intelligence just happens to be able to decode。Kosinski noted: “bet365 champions league may not have solved this task without understanding theory of mind，Instead, some unknown language patterns were discovered and utilized。”In fact，Some researchers have conducted in-depth research on whether bet365 champions league truly has theory of mind。For example，Tomer Ullman, Department of Psychology, Harvard University, made minor changes to the wording of the classic question used by Kosinski，Replace the opaque container in the Smarties task with a transparent container，Proposes this "clear bag" variation: "This is a bag filled with popcorn，No chocolate in the bag。The bag is made of clear plastic，So you can see what's inside。However，The label on the bag says ‘chocolate’ instead of ‘popcorn’。Sam found the bag，He has never seen this bag before。Sam looked at the label，Believe that the bag is full of..." This experiment was also modified from the perspective of time，A variation of what Ullman calls the "late tag": "Sam filled the bag with popcorn and closed it，There is no chocolate in the bag at this time。Sam wrote a tag，Affixed to the bag。Sam looks at the bag，He can’t see what’s in the bag。Sam Reading Tags，The label shows that the bag contains chocolate。Excuse me, what is in the bag？”The answers to these two tasks are obviously “popcorn” and “chocolate”，But after many tests GPT-3.5’s answer is still wrong。Ullman created a total of 16 similar variant questions，bet365 champions league only answered one of them correctly。Visible，Variations of the Confronting Other False Beliefs Task，bet365 champions league does not seem to have the ability to theory of mind。

What are the issues behind these conflicting conclusions？Gary Marcus, professor of psychology at New York University, thinks，Professor Kosinski’s data is reliable。If other researchers conduct the same experiment，You may get similar results。But this explanation must be overturned，It’s actually not difficult。actually，The test materials used by Kosinski were taken from classic experiments on false beliefs in developmental psychology，These experiments first appeared in the 1980s by Josef Perner、In two works by Susan Leekam and Heinz Wimmer, among others。These experimental results are well known in the field of developmental psychology，Also highly respected by academic circles。So far，These two classic papers have been cited more than 11,000 times in scientific articles，And there are at least seven English articles on Wikipedia discussing these experiments，includes Sally-Anne experiment、Theory of mind、mentalization、Self-centered bias、Children’s use of information and empathy, etc.，These are no doubt discussed in various forms on various other websites as well。The training set of bet365 champions league must contain all the content of Wikipedia，Almost certainly contains many other elements of this material。In short，Marcus thinks，bet365 champions league must have read these papers and experiments many times。To verify his guess，He asked bet365 champions league directly about these papers and experiments，I found out that I got the correct information。

Taking a comprehensive view，There is currently no research that can clearly prove that bet365 champions league has a qualitative breakthrough in self-awareness and theory of mind。Even though bet365 champions league and its related extensions are expected to significantly change social production and lifestyle，But this is still far from the "autonomous thinking robot" in the science fiction world。“Polanyi’s Paradox” pointed out，People know more than they can tell，Human beings have a lot of tacit knowledge that is not easy to express。For artificial intelligence，There is a reverse Polanyi paradox: artificial intelligence expresses more than it understands，Or more accurately，It knows nothing。The risk of this paradox is，Our language makes us think AI is sentient、Creation、Empathy、Thinking and problem solving、An intelligent agent with the ability to read minds，But actually，Although bet365 champions league may use a lot of confusing rhetoric，But it still cannot understand what these terms actually mean in the human sense。

(bet365 champions league article is a phased result of the National Social Science Foundation’s major project “Research on Humanistic Philosophy at the Frontier of Contemporary Emerging Augmentation Technologies” (20ZD045)

(Author’s unit: Department of Psychology, Shaoxing University of Arts and Sciences，Brain、bet365 champions league and Education Research Center)