ChatGPT is fast gaining popularity around the world. I recently had a chance to learn more about this revolutionary artificial intelligence (AI) application when I met Junichi Ito. A prominent entrepreneur and former director of the MIT Media Lab popularly known as the “God of IT,” Ito is now Director of the Center for Radical Transformation (CRT) at the Chiba Institute of Technology (CIT), where his students use ChatGPT in his classes. On April 10, Ito introduced Sam Altman, CEO of OpenAI which developed ChatGPT, to Prime Minister Fumio Kishida.

　　　　　“It’s just been a week since my students started using the application, but I am counting on ChatGPT to definitely develop their capabilities dramatically,” Ito remarked with a big smile.

　　　　　ChatGPT stands for “Chat Generative Pre-trained Transformer.” “Transformer” is a type of AI, an algorithm. Appearing on my “Genron” Friday Internet TV news show on April 21, Ito had this to say about ChatGPT:

　　　“ChatGPT is an AI chatbot built on a family of large language models (LLMs). The LLMs’ functions are easier to come to grips with if one thinks in terms of kana (phonetic alphabet) to kanji (Chinese ideographs) conversion in a word processor. When converting a kana, a string of kanji will appear on the screen. They are the results of the device having statistically learned what kanji follow what words after being trained on numerous Japanese sentences. The device is not errorless and the user may not get the right kanji at first, but eventually will by continuing to hit the key. GPT is very similar in function.”

　　　　　In other words, Ito said that, in interacting with humans, GPT functions as an instrument that statistically figures out what appropriate sentences should follow those before them. Ito has an image of GPT as continuing to select pertinent sentences in accordance with a certain algorithm rather than trying to attempt human-like conversations.

　　　　　As another image that he suggests would encourage a ready understanding of GPT, Ito cites the example of a huge library complete with all the books from around the world, explaining:

　　　　　“When you tell a librarian you are interested in learning about a specific subject, he or she is expected to name a few books. It isn’t that the librarian knows the answers to your questions but would find books that would show the way to the answers you are looking for. That is similar to how GPT functions.”
Neither the librarian nor AI knows the answer.

China and Russia Vigorously Utilizing GPT

　　　　　“Allow me to also note that AI really knows nothing about gravity, or mathematics, for instance,” Ito continued. But in math, when it comes across the simple ‘2+2=4’ formula, AI instantly recalls having seen it somewhere, and goes through the process of presuming that a ‘1+1=2’ formula it had previously stumbled upon would perhaps be related. So there is considerable ambiguity about AI, causing it to make mistakes rather unabashedly. It’s because AI figures things out on the basis of imagination only.

　　　　　“LLMs aren’t based on tightly-formulated structural rules. In human society, we sometimes meet those who simply recite somebody else’s remarks as their own without understanding them themselves. They just repeat what somebody important has to say.” Ito added with a chuckle: “AI is like such characters.”

　　　　　An ordinary citizen who is neither a programmer nor an AI expert can now access this enormously knowledgeable AI by the name of ChatGPT—free of charge, no less. One can handle GPT without being able to write programs; with ChatGPT anyone can be a programmer. But Elon Musk, co-founder of OpenAI, the firm behind the development of ChatGPT, has expressed reservations about the advent of the “age of dreams.” Ito explained:

　　　　　“What Musk is worried about is that, if the brain part of ChatGPT is to be opened to the public, criminals could exploit it. There have also been debates on whether specific people or teams have the right to decide what part of GPT is good—or not good—to make public. Musk himself has called for a six-month suspension in GPT development, although he has actually already founded a new firm to develop a new GPT model.

　　　　　Even if some countries refuse to use ChatGPT, China and Russia will utilize it vigorously. Ito maintains that if Japan is unable to use or understand tools that hostile nations have, we will end up being a dismal loser in all respects.

　　　　　ChatGPT, which got off to a good start last fall, tends to generate strange answers to questions, revealing some immature aspects at the present stage of its evolution. But there is no question it is progressing by leaps and bounds.

　　　　　To promote a further advancement of ChatGPT, humans can supply additional information while setting strict standards of right and wrong, Ito emphasizes. He asserts that humans will be able to control ChatGPT’s thought system by injecting values, ethics, and philosophies and further adding information and wisdom required for its maturity.

　　　　　If humans are to make the final decision, how then should we determine the relationship between ChatGPT and those with evil minds in this world? Russia’s war of aggression in Ukraine has led to production of a succession of AI weapons with increasing lethality. Ito observed:

　　　　　“Good and evil in the world exist almost to the same extent. Furthermore, humans cannot control each and every happening in the world. That is why we must attach importance to governance and the role played by the media. AI, including ChatGPT, provides individuals with tremendous energy, like attaching a jet engine to a human body. That means the ability to generate awesome energy in good or bad directions. This makes all the more important who controls AI and ChatGPT. The state controls them in China and Russia, while private enterprises do in the US. Europe is hesitant, and Japan finds itself halfway between the US and Europe. But I am firmly convinced that Japan will sooner than later be able to play quite an important role in this area.”

Japanese Capable of Mastering AI

　　　　　Japan got off to a belated start in information technology (IT). But Ito stresses that the Japanese are equipped probably with the best mindset to master it for the maximum benefit of mankind. He noted:

　　　　　“I am of the opinion that in the US and Europea, where Judeo-Christian values prevail, humans are positioned above animals, things, and nature. Seemingly viewing AI and robots as entities ready to oppress humans, Americans and Europeans tend to feel emotionally and religiously uncomfortable with AI and robots. Meanwhile, Japan is a nation of the polytheistic Shintoism, where humans are viewed as part of nature. I suspect that the Japanese people may generally feel nothing very strange about robots and machines having souls like ours. I therefore feel that we Japanese are philosophically and religiously better suited for AI utilization. In fact, I would think it quite possible for Japan to see AI and society fuse together well to sustain a humane governance of AI.”

　　　　　Ito has invited Tenzin Priyadarshi, President and CEO of the Dalai Lama Center for Ethics and Transformative Values at MIT, to CIT to conduct classes jointly. The duo is teaching philosophy using the “Socratic Method,” asking students questions and encouraging them to think and learn rather than merely listening to lectures. Ito had this to say about this approach:

　　　　　“Against all expectations, ChatGPT has proved to be extremely good at the Socratic Method of thinking philosophically. I personally am learning much from asking ChatGPT a lot of questions, drawing pertinent answers, and deepening my own thinking by scrutinizing them.”

　　　　　Ito is confident about his ongoing teaching method and thinking about philosophy with a Tibetan lama at a science university, remarking:

　　　　　“Human society is filled with half right and half wrong, and we wish to lead it in the right direction. I believe we have nothing to fear from AI or ChatGPT, if we can implant in AI the indigenous Japanese philosophy and ethnics of ‘wa’ (harmony) to that end.”

　　　　　A program designed to realize this purpose is slated to be announced as early as mid-June this year. I am eagerly looking forward to it. (The End)
(Translated from “Renaissance Japan” column no. 1,047 in the May 4～11, 2023 combined issue of The Weekly Shincho)