In this blog post, we analyze the thinking ability of AI from the perspective of contradiction and explore the nature of thinking.
AI, talking like a human
“Hello, sir.”
This is how JARVIS, the AI assistant of Tony Stark, a character in the globally popular superhero movies Iron Man and The Avengers, greets his master Tony Stark. In the movie, Tony Stark says to JARVIS, “A little ostentatious, don’t you think?” and JARVIS responds to an enemy trying to control him by saying, “I believe your intentions to be hostile.” In both movies, the AI JARVIS is portrayed as if it can “think” and converse with humans as an equal.
The “thinking” of humans and AI
Long ago, Descartes said, “I think, therefore I am.” The answer to the question, “Can humans think?” is commonly “Yes, humans can think” for most people, and it is difficult to find anyone who would argue otherwise. As a human being, I am able to ask myself the question, “Can I think?” because I am capable of thinking in the first place.
If so, can artificial intelligence think? Even if we reserve judgment on JARVIS in science fiction movies because we do not know the specific level of science and technology applied to it or the limits of its capabilities, it remains questionable whether artificial intelligence at its current level is capable of the “thinking” mentioned above. Furthermore, what are the characteristics of “thinking” that distinguish it from other similar behaviors? What is the difference between humans and AI in terms of “thinking”?
To gain a deeper understanding of these issues, we will introduce two experiments on AI and thinking: the Turing test and the Chinese room. The Turing test is an experiment proposed by Alan Turing in 1950. Based on the belief that “if a computer’s response to certain inputs cannot be distinguished from a human response (specifically, if the computer deceives the experimenter 30% of the time during the entire test), then the computer can be considered intelligent and capable of thinking,” this test examines how similar a computer’s responses are to those of humans.
When I first heard about Turing’s belief behind this experiment, I had one question: “If a computer (or artificial intelligence) has a vast or high-quality database and simply compares and contrasts information according to algorithms without understanding the input, should that be considered ‘thinking’?” John Searle, who had the same question as me, devised a thought experiment called “The Chinese Room” to refute Turing’s belief.
The thought experiment is as follows. First, a person who does not know any Chinese but can distinguish the appearance of Chinese characters is placed in a room with two windows, one for receiving questions and one for receiving answers. The person is given a list of pre-prepared Chinese questions and answers. An observer outside the room, who does not know that the person in the room cannot speak Chinese, observes the person in the room answering the Chinese questions.
To the observer outside the room, it would appear that the person inside the room understands all the Chinese questions and is responding appropriately. However, in reality, the person inside the room is simply responding according to the list, rather than understanding the Chinese questions and thinking through the answers. Based on this, John Searle concluded that the Turing test cannot be used to determine whether artificial intelligence has actual intelligence and is capable of thinking.
Criteria for “thinking” – Focusing on “thesis, antithesis, and synthesis”
Like calculators and search engines, and like the person in the Chinese room, “thinking” can be said to involve a process of information processing in that it involves opening and searching a database of experiences and knowledge when input is received and then producing a response. Therefore, the question of artificial intelligence and thinking ultimately boils down to the question of what criteria distinguish “simple information processing” from “thinking.” The question is what elements must be included and what must be possible in order for something to be called “thinking.”
Of course, as many philosophers and engineers have failed to find criteria that satisfy everyone, it is not easy to establish strict criteria. However, in this article, I would like to propose one of the criteria for “thinking” mentioned above, which is an element that “information processing” must include in order to go beyond itself and enter the stage of “thinking.” It is the following short question.
Can the process of “thesis, antithesis, synthesis” be carried out?
“Thesis, antithesis, synthesis” is the three stages of Hegel’s dialectic, consisting of “thesis,” “antithesis,” and “synthesis.” Simply put, “thesis” is a specific proposition (or argument) that exists with opposing sides. “Antithesis” is another proposition that is opposite or contradictory to the preceding ‘thesis.’ When ‘thesis’ and ‘antithesis’ meet, because they are two contradictory propositions, they clash and connect over a much longer period of time than when two similar or unrelated propositions meet, producing numerous corollary propositions and secondary knowledge in a ” This “productive logical process” repeats itself, and the original “jeong” and “ban” are integrated into a single proposition, “hap,” which is deeper and greater than the original “jeong” and “ban.” The resulting “hap” is qualitatively more developed than the existing “jeong” and “ban” and has the property of being applicable to all situations related to the subordinate propositions “jeong” and “ban.” It does not end there. “Synthesis” becomes a new “thesis” and goes through the logical process of encountering contradictory “antithesis” again, aiming for an absolute truth that can be applied to various situations. Therefore, thesis-antithesis-synthesis is a logical process with a single completeness that leads to an absolute truth.
In other words, proposing “thesis-antithesis-synthesis” as a standard for “thinking” means the following: When a thinking being has a certain proposition (in this case, information) and a proposition (information) that contradicts it, it should not simply store and list them in a database and compare them, but should carry out the process of thesis-antithesis-synthesis using the existing proposition (information) and the new proposition (information) as materials. Furthermore, through the process of thesis, antithesis, and synthesis, the newly created “synthesis” must be included in one’s database to create a qualitatively improved database, and when an input is received, the database must be searched to produce a result.
In my opinion, “thinking” is not a process of listing all new propositions (information) in a huge database, searching for them one by one each time a response is required, calculating and comparing them, and selecting the best one to submit as output (this is the process by which a PC stores everything in memory and selects the output). “Thinking” is the process of colliding, connecting, and integrating elements within the database to be searched, thereby improving the quality of the database and producing efficient output. Furthermore, it can be said that ‘output development’ should come from the development of the database itself, rather than the development of ‘output selection’ and ‘submission’ (e.g., faster simple calculation speeds of computers).
The reason why I believe that synthesis can improve the quality of a database can be explained schematically as follows. Suppose that there is “contradictory” information A, B, and C as candidates for output for input P. Even if A is submitted through calculation for input Q similar to P, there is no guarantee that A will be the appropriate output if information C is added or if P and Q are similar but not completely identical. Therefore, the operation must be repeated every time in order to submit the output for not only P and Q, but also all similar inputs. However, if A and B are integrated into D and C and D are integrated into E through the process of synthesis and contradiction, E will be the appropriate output for both P and Q and similar inputs. Therefore, the quality of the database itself can be improved through the process of synthesis and contradiction.
Does artificial intelligence “think”?
Let’s connect the criteria I proposed, that is, the ability to perform the process of synthesis, to the case of AI thinking, which we were curious about. Considering the current level of AI, such as Siri on the iPhone and AlphaGo, which defeated Lee Sedol, it is clear that Siri, which responds to a few limited questions with predetermined patterns, is incapable of performing the process of synthesis.
AlphaGo, which defied expectations and won a game of Go in March 2016, a game with an almost infinite number of possible moves, is called a “concentration of deep learning” technology. However, the core of deep learning is not “development” through the process of synthesis and opposition, but the “classification” of vast amounts of data. AlphaGo learned through prediction based on data clustering and classification using its unimaginable computing power and technology, and found the optimal output by calculating all the possible moves on a 19×19 Go board for each situation, thereby defeating Lee Sedol. However, it did not develop itself toward something absolute.
In other words, I believe it is unreasonable to say that AI at its current level is capable of thinking. AI is still only capable of pretending to be human by listing, contrasting, and comparing data based on its superior computational abilities, in other words, by giving answers similar to those of humans. It is unreasonable to define AI, which does not develop through a logical process of thesis, antithesis, and synthesis, as capable of thinking with self-awareness.
Conclusion
Even as you read this article, science and technology are advancing. Thanks to this, the output of artificial intelligence in response to input is becoming more and more human-like. Perhaps someday we will be able to talk about artificial intelligence and Goethe or Nietzsche. However, even if artificial intelligence can talk about Goethe or Nietzsche, “thinking” must include the possibility of self-development (in a narrow sense, the development of a database) through the process of thesis, antithesis, and synthesis.
