Reporter | Yin Qinglu
Edit | Yellow Moon
ChatGPT is considered by many to open the door to general artificial intelligence (AGI). AGI's ability to have the same intelligence as humans, and possibly even surpass humans, was once a staple of science fiction, and now people's joy and fear seem to have come true.
On the eve of this storm, the novel author Mu Ming launched his first collection of Chinese short stories "Like a Circle", from ancient to today's neural networks, and then to the near-future AI to surpass human beings, Mu Ming made his own analysis and thinking on the relationship between man and machine.
Mu Ming graduated from the Department of Intelligent Science of Peking University as a Google programmer, and began writing short and medium-length science fiction novels in 2016. Combining her science and engineering background and science fiction writing, how does she see the boom in machine writing and chatting? And how to respond to technology-related controversies? In this interview, starting with ChatGPT, she talks about the possibilities and costs of technology and "what machines can't write for the time being"—which is what Mu Ming is doing: weaving multi-layered stories and thereby restoring human complexity.
Recent photo of Mu Ming
01 Technology talk: Why can't people learn through the feedback of machines?
The first story in the collection, "Casting Dreams", is about the history of machine learning. Mu Ming used the method of historical reconstruction to let artificial intelligence possess the pre-Qin era apparatus. The novel features two teenagers making dolls as the main storyline, pointing out that machine learning was originally based on rules, but due to the large number of rules, this method quickly became limited, and later, artificial intelligence entered another path, that is, bottom-up, reinforcement learning in reward and punishment and feedback (the large language model based on ChatGPT is also trained through reinforcement learning). However, humans can't iterate as quickly as machines, and are accustomed to top-down "teaching and learning." The differences in learning styles seem to doom humans and machines to understand each other.
In the following long story, Mu Ming wanted to propose the concept of "phase change point". She explained the concept as follows: "The coupling relationship between human beings, language and information has never been as close as it is today, just as water will boil when it warms up to 100 degrees and change from liquid to gaseous state. Someone is convicted for speaking on social networks, and amateurs who master the speaking skills can become big Vs, this is all happening. "Internet information is built on the sum of human minds from ancient times to the present, so in order to deal with the latest AI, we must also learn the essence of human knowledge and dig out the modernity in it."
Interface culture: Many people think ChatGPT opens up the future of general artificial intelligence, but science fiction writer Ted Jiang points out that it's more like "compression software" that preserves obfuscated text. Your writing is heavily influenced by Ted Jiang, so what would you think of his views?
Mu Ming: I found that in the recent technological boom, the Chinese Internet fever is even a little higher than that of the English-speaking world, which is a phenomenon worth observing. In the high heat, there are two voices, one is very optimistic, many domestic companies have said to benchmark the ChatGPT development model. The other is professionals like Ted Ginger or Greg Egan, who will be wary.
In my opinion, the first thing to pay attention to is what the position of the speaker is. Ted Jiang has this view because his expectations for ChatGPT are a large and comprehensive expert system, and many of my research friends also want it to give a professional compilation of data, so they will care whether it is perfect lossless compression or flawed lossy compression. But if you think of it as an ordinary person who is intellectually similar to yourself and can help inspire, then the ambiguity of ChatGPT becomes an advantage.
Science fiction writer Ted Jiang. Image source: Wikipedia
In my opinion, "compression software" is not a very strong criticism. We should take a step back and consider the existence of people, human learning is also a compression process, whether it is reading books or listening to lectures, the transmission of information must be lost. Some new media articles are well written, but they have to take eye-catching titles, and Ted Jiang's article is written in thousands of words, with very sophisticated arguments and logical deductions, but readers receive only a single point of view. Therefore, Zen Buddhism will say "no writing", and the ancients have long had this awareness.
Interface culture: Some experts pointed out that the expression of artificial intelligence cannot go beyond the existing database, and there is no real originality, Han Shaogong's article "When Robots Establish a Writers Association" mentioned that the master-slave relationship between man and machine is difficult to change. However, it has also been seen that machine writing has impacted the humanities, and some universities abroad and Hong Kong have banned ChatGPT to assist in writing papers. Why do we think that machine writing cannot surpass humans, but also worry about its impact?
Mu Ming: It is difficult to discuss human beings as a whole, and many experts often compare the highest wisdom of humans with machines. As a new author who is still learning to write, I am more sensitive to this kind of expression, after all, the learning process requires a lot of imitation practice, and the so-called innovation often comes from the re-integration of existing knowledge. Writer Wei Zhou said: "I will not be replaced by AI, I am AI." "I said something similar. When we put the position a little lower, we feel that ChatGPT is quite kind, like a child with strong learning ability but limitations, and will look forward to its performance later.
For me, the best thing about ChatGPT is that innovation is a vague concept, and it may be possible to suggest a reference line for innovation, that is, to quantify the value of the fruits of human mental labor. For example, a new novelist, we can let ChatGPT to see if he is easy to imitate, and we can also use ChatGPT as a unit of measurement to see whether this person's creation is 0.8 times ChatGPT or 1.5 times, which are very interesting application scenarios.
Moreover, artificial intelligence does not appear out of thin air, its iteration itself has the help of humans, since machines can learn through humans, why can't people learn through machine feedback? Twenty years ago, working in writing also required consulting library materials, and now search engines have become a standard skill. As Edward Ashford Lee (computer scientist, author of Plato and the Tech Nerd and Co-Evolution: The Future of Human-Machine Convergence) points out, technology and people evolve co-evolved, and artificial intelligence is not just a tool, but has largely shaped who we are today.
Co-evolution: The Future of Human-Machine Integration
Edward Ashford Lee, translated by Li Yang
CITIC Press, Synops Culture 2022-6
The key to the impact of the humanities is that machines redefine what issues deserve more attention and which productivity they liberate. Domestic humanities have always paid attention to the crawling and combing of data, but when machines can replace many jobs, we must examine whether we should pay attention to this kind of work. Literary writers usually value language, but when machines learn language styles quickly, I wonder how to write things that machines can't write for the time being, such as texts that are more logical and have a lot of hidden content.
Interface culture: In the discussion of ChatGPT, many people are optimistic about the technological leap; On the other hand, OpenAI has been exposed to hiring outsourced sweatworkers in Africa, who are subjected to a lot of abuse and murder every day. You've also written in your novels about people who are discriminated against without high-tech glasses. What would you think of the contradictions of this technology?
Mu Ming: As far as my industry and science fiction circle are concerned, most people are optimistic, and practitioners with deep humanities backgrounds will emphasize the alienation of technology to people, and film and television dramas such as "Black Mirror" and "Life Cutting" are also critical of high technology - fanaticism and resistance exist in a society at the same time, which is very interesting. When you don't understand the details of technology, it's easy to form a tendency, and I myself have two backgrounds and am willing to use positive and negative thinking to deduce, so I will provide two POVs (perspectives) when writing novels. Specifically, my novels deal with the question of "possibility" and "cost," both of which are indispensable.
This also responds to why I write ancient stories. Now when it comes to "technology" is information technology since the 20th century, in fact, the word refers to "technology", and craftsmanship is also technology. And I wanted to examine the technological changes of the time—of course, what we now call historical events—and examine how technology impacted the people involved in that context, and how people responded to change. The story of "Casting Dreams" is obvious, it tells the process from no bronze to mass production of bronze. Zhang Guangzhi wrote in "Art, Mythology and Sacrifice" that the emergence of bronze and the centralization of power were accompanied by each other, because casting required a lot of manpower and fine division of labor, as well as foremen to manage, only then there would be tribal annexation into feudal society transformation, and the development of technology produced weapons, further consolidating rule. When we put technology into history, maybe there will be such a strong panic about technology. In fact, in a sense, the emergence of words is much more terrifying than artificial intelligence, and people who master words once felt that they mastered the mysteries of the world, but at least people who use artificial intelligence will no longer think this way.
"Like a Circle"
By Mu Ming
Single Reading, Casting Culture/Shanghai Literature and Art Publishing House, 2023-2
02 Creation Talk: What works can be written by humans but not by machines?
What can a human write but a machine cannot? In Mu Ming's view, when artificial intelligence pulls the proud learning behavior of human beings off the altar, what writers need more is the ability to deeply understand social mechanisms, organically organize information, and find the close relationship between everything under the divided system. She mentioned in her preface that in the era of information explosion, the overall narrative similar to "Dream of Red Mansions" is missing. Quoting Tokarczuk's Nobel Prize speech, she said: "Literature is not ready to tell the future, to tell about the rapid transformation of the world—lacking metaphors, perspectives, and new allegories." ”
The novel of the same name, "The Ring of Wan", is such a story about complexity. A pair of gentry father and daughter in the late Ming Dynasty accidentally encountered the "Wan Rotation Ring", the inner wall of the jade ring was painted with subtle mountains and rivers, and it was rotated for half a circle, apparently an ancient reproduction of the Mobius ring. The Mobius ring realizes the transfer of time and space in different dimensions, and also corresponds to the "method of emptiness" of the ancients in Chinese gardens, through the skillful arrangement of scenery, a garden can also accommodate vast lakes and seas.
Interface culture: Your disciplinary background is intelligent science, and you are also engaged in related industries, why would you write science fiction outside of this? How is it different from practical work?
Mu Ming: At the beginning, I definitely wanted to integrate my two knowledge systems, on the one hand, the work experience of science and engineering technical background, on the other hand, as a long-term reader's extensive reading experience, I found that the expression system of these two worlds in contemporary life is divided. British scientist C.P. Snow put forward the concept of "two cultures" in 1969, he believed that due to the high degree of division of modern disciplines, people who understand Shakespeare do not understand Newton's fundamental laws, so I chose science fiction, a novel with a strong rational color, as a medium for processing and presenting multi-dimensional information.
My novels often use "confluence" as a starting point for conception, which is also a technique I learned from Liu Yukun's short stories. In my novels, the core technical concepts and deduction ideas are often contemporary and imported, but the cultural background, perspective, characters, ideas, etc. are non-contemporary and Chinese. For example, "Wan Zhuanhuan" can be regarded as the late Ming version of "Interstellar", "Casting Dreams" can be regarded as the pre-Qin version of "Westworld" (seasons 1 to 2), and "Fake Hands in People" is a new interpretation of "Broken Soul Gun" or "Ordinary People" under the recent future environment and technological progress.
However, the more I write, the more I find that in addition to "building bridges" between fields, this way of writing has many functions, and sometimes it can be well reflected on just a social deduction of the technology itself. For example, I wrote "Who Can Own the Moon" last year, in which 3D modeling and NFT are particularly close technologies that have penetrated into the daily lives of ordinary people.
Interface Culture: The story of "Wan Turn" embodies what you said about merging and bridging the two worlds. How did you come up with this story?
Mu Ming: When I was reading the history of the Ming Dynasty, I found the documentary "600 Years of Kunqu Opera", one of which is about Qi Biaojia, a political figure in the late Ming Dynasty, who seems to be a serious political leader, but he is also a good expert in opera and gardening, but our history will not deal with this. This allowed me to see the complexity of people, and the similarities between people and stories, because stories are ambiguous and can present the structure of the world on multiple levels.
I think that the popular "dimensionality reduction strike" is too simple and rude, which is a simplification of the real world, and literary and artistic works should do the work of "upgrading", not only from two-dimensional to three-dimensional space, but also by weaving stories to restore the complexity of people. "Wanzhuan Circle" is a cute little craft that can be enjoyed repeatedly, and has a lot of things inside, which is also the positioning of self-reference and expectation of this book. In my opinion, whether we can resist "re-reading" is an important criterion for judging literary works and the thinking ability of authors, and whether we can attract potential readers from the outside (shallow characteristics), and whether we can lift the weight lightly, arrange big problems and big strokes in a small space, and pass them on to those who are destined for it, is the embodiment of "skill" or "ingenuity".
Ming Wen Zheng Ming Lanting Ya Collection Scroll
03 Women talk: Science fiction needs female power, and reflective technology needs a female dimension
When thinking about "can machines write", we tend to substitute the highest level of human beings, which includes masculine and elitist thinking. In Mu Ming's view, when the emerging technology industry has an impact on those who have achieved fame, women can embrace new technologies without caring so much, feeling that "this thing is very good, similar to us".
The short story "Who Can Own the Moon" tells the story of a woman in the modeling industry, He Xiaolin lived a poor life with his mother since he was a child, but by chance learned modeling technology and began to mint NFT artworks. The article writes that non-fungible tokens satisfy people's desire for self-expression and ownership, but He Xiaolin is the first artist to use CC0 Creative Commons licenses, which means she gives up all personal rights. Rather than owning, "it's not owning that makes her work take on rich forms." ”
At the end of the interview, we talked about how the gender dimension can help us reflect on the changes in technology and inspire women creators to seek change in the current era.
Interface Culture: How can science fiction help us rethink the relationship between man and machine?
Mu Ming: Science fiction has been talking about the relationship between man and machine from the beginning, such as Mary Shelley's Frankenstein, which deals with the relationship between "creation and creation". The second level is "controlling and being controlled", through which many dystopian works introduce the discussion of power and inequality. Finally, people and artificial intelligence also have a layer of mutual mirroring relationship, people to understand themselves reflexively through artificial intelligence.
Stories are a great genre to tell these relationships all. There are many classics that address each of the three themes, and my work attempts to explore all three at the same time. Artificial intelligence is also a topic with a very wide imagination space, in addition to the strong artificial intelligence (AGI) common in science fiction works, the technology with artificial intelligence components is used every day, it is very close to life, not as abstract as theoretical physics, and writers can use a lot of materials. Some national writers write about artificial intelligence, and their works also have distinct personal characteristics, such as McEwan's "Machine Like Me" to discuss Britain's situation in globalization, and Ishiguro's "Clark and the Sun" is more emotionally deep.
Of course, you can also see their position and limitations, as they did with Ted Jiang. Both writers are men, so in McEwan's writing, the omnipotent artificial intelligence must be robust male, and the artificial intelligence written by Kazuo Ishiguro is a little girl with flawed abilities. So, each of us can think about what is your relationship with artificial intelligence, are you satisfied with the existing narrative, satisfied with the views of well-known writers?
Interface culture: There are often female characters in your novels, and the development of technology has increased the subjectivity of women, but sometimes it also exacerbates the exploitation of women, which is another contradiction. In your opinion, how can the gender dimension help us reflect on technological change?
Mu Ming: In fact, "the obscured female narrative tradition in civilization" is one of the skeletons of the book. "Self-Order: From Ape to God" repeatedly suggests that women, as mothers, are the earliest narrators, both in individual experience and in group history. "Telling" was the first and most important power of ancient humans. But since the advent of agriculture and the age of tools, the main form of power has shifted from mind and spirit to physical ability, and this transformation is the result of technological evolution, which to some extent promoted the emergence of patriarchy and private ownership, women gradually lost their power, and "telling" was monopolized by men. This theory appears in the discourses of Engels, Beauvoir and others.
In "Casting Dreams", the female narrator has nothing but stories. They cannot resist the times, and can only silently wait for the inheritance of those who are destined. In addition to the various meanings mentioned above, "ring" is also a symbol of inheritance. The first time, Natsuhime gave Fuji, the second time was given to the dolls created by Fuji, and the third time was Zigang in "Wan Zhuanhuan" to Zou'er. "Inheritance" is an implicit theme in this book, whether it is the traditional relationship between master and apprentice, father and son, father and daughter, or the soulmate, listener/viewer, and even the relationship between the creator and the created. This process, despite all the problems it brings, unleashes to some extent the abilities of female/weak people who have been imprisoned since the Stone Age.
So, when will the era when "storytelling" becomes important again, and women and even human beings are no longer limited by the shackles of physical and cultural construction? It is now and in the near future. Bronze downgrades women, machines upgrade women. Under such a picture, there is a speculation that women/weaklings in the traditional sense will return to center stage. The "weak" here actually also means "literature belongs to the weak" that Takeuchi said.
This is worth thinking about female creators, the era we are in is very critical, as the first generation of only children, well educated, and have done a lot of work, but not seen much, including the country is also recently published some female science fiction author anthology, I did not have a strong sense of gender consciousness when I started writing. And literature, like bottom-up machine learning, can solve many problems that condescending, grand narratives cannot. This is one of the reasons why I insist on creating, and I hope to see more female creators working together.