Written by | Weng Yuqing
Edit | Yellow Moon
On October 31 of this year, Collins Dictionary announced that "artificial intelligence" (AI) was named the word of the year. AI is defined in the dictionary as "the simulation of human mental functions by computer programs". According to Collins, the use of the word has quadrupled in the past year.
Led by ChatGPT launched by OpenAI, various generative AI products have emerged in an endless stream over the past year, and a whole set of thriving ecosystems have been formed so far. Leap Motion's Midjourney, Google's MusicLM, and Microsoft's Copilot. It was as if we had stepped into the world of science fiction overnight. In the "Iron Man" movie, Jarvis, who can do everything to build armor, manage satellites, protect nuclear bomb codes, take good care of Stark's food, clothing, housing and transportation, and say a few witty words from time to time, seems to be no longer far away from us.
But AI brings more than just convenience. Not to imagine the "Skynet" system that cleans up human beings, artificial intelligence at today's level of science and technology has brought a lot of impact to society. AI uses fake artworks to make us think about the extension and connotation of art, and torture what is "fake", what is "real", what is intelligence and creativity. Unlike the impact of the first and second industrial revolutions on the manufacturing industry, the cultural industry was the first to be affected. As the emergence of artificial intelligence has blurred the concepts of "originality", "artistic expression" and "fair use", the legislation related to "intellectual property" needs to be revised and supplemented urgently. In addition, the proliferation of AI tools has also caused widespread concern among white-collar workers, and this disruptive technology is bound to have a profound impact on the already skewed labor relations.
There's an analogy that says people look at emerging technologies like they see a train coming from afar. When you are far away, the speed seems to be very slow, and you can't always reach the front of you. When the train really arrived, it whistled by, but we didn't have time to react. Maybe now is the moment when the AI train is coming. As individuals and as a community, we have to be prepared. As Yu Mingfeng, associate professor and deputy director of the Department of Philosophy at the School of Humanities at Tongji University, said in an interview with Interface Culture: "What does the widespread application of artificial intelligence mean? What are the profound challenges it will bring to the economy, politics, daily life, and even human self-understanding, and to our philosophical thinking, basic values and beliefs, social laws, etc.? This is what will happen in the future, and it is also the more important reason why we as thinkers pay so much attention to it." So, a public discussion on this is very necessary. ”
01 AI and Creation: Copying or Creating, Learning Xi or Robbing?
Keywords: Copyright, ChatGPT, Midjourney, Science Fiction, Music
On November 27, the Beijing Internet Court made a first-instance judgment on the country's first case of copyright infringement dispute over AI-generated images, ruling that the plaintiff was entitled to copyright in accordance with the law because of activities such as designing prompts and modifying parameters that constituted "intellectual investment" and "personalized expression". This case is a milestone for the future application of AI in the cultural industry. According to an interview with industry insiders, "its significance lies in the legal recognition of the originality of AIGC's works (referring to Artificial Intelligence Generative Content) and the right to claim damages for infringement." This is an important manifestation of legislators' enthusiasm to keep up with the pace of technological development. ”
In determining whether something is a "work", the law mainly considers several factors: whether it belongs to the fields of literature, science and art, whether it is original, whether it has a certain form of expression, and whether it is an intellectual achievement. In this case, the court held that the AI-generated image in question was undoubtedly an artistic expression, because it was no different from the human-created images that we usually see, and that the plaintiff's input into the image generation process satisfied the elements of "intelligence" and "originality". Since AI currently does not have free will and is not a legal subject, it is still essentially a human using tools to create, so people enjoy the protection of copyright law.
"My painting was stolen?" The report "The Verdict of the First Case of Copyright Infringement of AI-Generated Images Released" pointed out that the core purpose of the copyright system is to encourage creativity, so when new technologies appear, the copyright law should be appropriately used to encourage more people to use the latest tools to create.
With the introduction of generative AI such as ChatGPT into people's lives, AI-generated works are forcing people to rethink what art is. This seismic change is similar to the first advent of photography, which has brought about the impact of painting as an art of "representation". Interface Culture once discussed the saying that "AI ends art", and pointed out that it is not art itself that is the end of emerging technologies, but a specific art narrative. After the popularization of photographic technology, modernist art found personal emotional expression and flatness as an anchor, expanding the extension of art, leaving a number of unique masterpieces for future generations. Today, AI itself is an extremely provocative work of art. With the lowering of the threshold for creation, the coexistence of controllability and randomness of generated works, and the technology companies behind AI that conduct big data training, they have all broken through the boundaries of modern or postmodern art and pointed to a new kind of artistic life.
The impact of AI on creators on a global scale is not limited to the visual field. In February of this year, the famous science fiction magazine "Clarke's World" suspended the call for submissions because it received too many AI-generated works to review the manuscripts. However, Sheila Williams, the magazine's editor, also said that it is clear at a glance that the articles written by machines show that there is still a long way to go for AI to replace writers.
In terms of music, in May this year, AI Yanzi Sun and artificial intelligence covers attracted attention. Li Xingyu, a musician and sound space designer, pointed out in an exclusive interview with Interface Culture that the application of AI in the music industry has long been very extensive. The essence of music is similar to mathematics or finance, it is a purely mental activity, and it is very easy to combine music with technology to produce qualitative changes compared to the creation of physical forms such as painting. But when technology is monopolized by big companies and subordinated to their profit-making goals, the living space of ordinary creators is squeezed. Big data analysis has given birth to a large number of Internet celebrity saliva songs that do not seek merit, but seek no fault, and the playlist culture has cultivated listeners into "children" who passively accept recommendations and do not know how to explore independently. Li Xingyu said pessimistically in an interview: "The most likely scenario is that the platform will distribute resources to its own AI, and at the same time, the head artist will still ask for his share, and in the end there will be nothing left for other creators." ”
The above cases involve the content output of AI, and at this stage, AI training is still inseparable from the input of human creativity. This year's leading generative AI relies on large language models, which require a large corpus to train. The ideal corpus isn't a one-size-fits-all — Russell Kaplan of Data Startup Scale AI points out that the ideal training target is a work that is long enough, well-written, and factually accurate. Many times, they are derived from literary works. The seizure of corpus by technology companies has allowed writers to start a battle to defend their rights, and it has also made readers or viewers think more deeply about concepts such as "fair use", "intellectual property", and "creativity".
In June, American writers Paul Tremblay and Mona Awad sued OpenAI, accusing the company of using their work to train its AI model, ChatGPT, without authorization, and seeking damages. The complaint points out that much of the data OpenAI used to train ChatGPT comes from copyrighted works and is most likely provided by pirate sites such as Library Genesis.
Some of the writers we know well have joined the showdown. In July, more than 10,000 writers from The Authors Guild, including Margaret Atwood and Philip Pullman, signed an open letter demanding that leaders at Open AI, Microsoft, Meta Platforms and others not use their work to train AI without permission or payment. In September, the association and George W. R· R. Martin and 13 other writers have filed a class action lawsuit against OpenAI. Writers argue that since their work is being used to train for-money AI products, they deserve to be compensated. They also expressed concern about the proliferation of parodies. The point is not just that AI infringement, but that the purpose of such infringement will seriously harm the rights and interests of writers, similar to the aforementioned music industry.
In summary, many creators are already actively responding to the emergence and development of AI and promoting the improvement of copyright law. But the broader concern is not that a single work is infringed, but rather the impact of "parody culture" on the creative industries as a whole. According to the article "Rethinking creativity: creative industries, AI and everyday creativity", the advent of AI has prompted us to rethink the concept of "creativity". From the perspective of intellectual property, creativity is regarded as a kind of capital, a fixed expression that can be clearly priced, and the labor dimension of creativity is often ignored. The generation of creativity first requires the creator to be able to meet the needs of food and clothing, and to put in a lot of labor to improve self-cultivation, so as to maintain creativity and make inspiration come at a certain moment, and these labor is often unpaid.
02 AI and Human Work: Replace, Pred, or Drive Change
Keywords: Hollywood strike, OpenAI outsourced workers, "post-work era", education industry, modeling industry
The Hollywood strike, which lasted for five months this year, is a counterattack by creative industry practitioners against AI from the perspective of labor relations. Interface Culture has reported that in addition to the issue of profit sharing of streaming platforms, AI is also an important issue in this strike. The Writers' Guild demands that AI can only be used as an aid tool for screenwriters, and the Screen Actors Guild also opposes AI-generated characters replacing various actors. Similar to previous writers' protests, the actors questioned "whether producers have the right to let AI Xi actors' work."
After a five-month stalemate, the producers finally made major concessions in November. According to The Guardian, the main wins for the writers this time included that studios were not allowed to allow AI to generate raw materials and then ask writers to revise them, nor could they allow AI to modify the writers' work. In an interview with The Guardian, MIT economist Simon Johnson said the agreement reached by the writers was "smart" because the writers could still use AI as a research tool. "AI is controlled by the writers, not the studio," Johnson said. It will not be an automated technology, but a complement to human work. ”
However, much of the success of Hollywood writers and actors is due to Hollywood's global reach and the base of their popularity. Not everyone is capable of rising up against big platforms and AI. Interface Culture has been concerned about the plight of AI taggers, with OpenAI hiring Kenyan outsourcers to moderate its content at $1.32 an hour. The Times' investigative report exposed the condition of these low-wage workers, who have long been responsible for flagging violent, pornographic, hateful data, immersed in impactful and filthy content, and many have suffered from psychological problems such as PTSD, depression, anxiety, insomnia and nightmares.
This year, Shanghai Translation Publishing House introduced the book "The Post-Work Era: Labor in the Era of Platform Capitalism", which deeply analyzes the "micro-work" system in the era of platform capitalism. It appears to be creating free, resilient, decent, geographically unconstrained new jobs that are a boon to the Global South, but it is a further distortion of the global job crisis. Amazon's micro-job site, Mechanical Turk, for example, offers a lot of data labeling jobs, but only a handful of people in developed countries can get paid by bank transfer, and workers from countries in the Global South can only receive gift cards to buy goods and services offered by Amazon. At the same time, due to the nature of piecework tasks, workers need to spend a lot of time looking for their next job, and the time to find a job is even longer than the time to complete the job. Not only is there no guarantee of rights, but there are serious ethical problems, as workers do not know what the data they themselves enter will ultimately be used for. The U.S. military and Google have co-developed an artificial intelligence program that allows data taggers to classify drone video, but not to help the military identify targets on the battlefield.
With the development of AI technology, even such precarious jobs are becoming difficult to maintain. According to the article "Chinese County Workers Training Artificial Intelligence", AI researchers have begun to try to "feed" unlabeled data and partially labeled data to machines, that is, "semi-supervised Xi", rather than relying on self-supervised Xi and data annotation of manual annotation. Tencent, Alibaba, ByteDance and other major companies are developing self-supervised Xi algorithms, and even some data annotation companies have 60% of the content from machine automatic annotation. Data annotators used to be called "teachers of artificial intelligence", and thousands of pull-out boxes per person every day taught artificial intelligence to understand the world, but they did not expect that "teachers" would be replaced by "students" in the blink of an eye.
Artificial intelligence is replacing more and more jobs. The Southern Weekly report shows the challenges of AI for the education industry and the modeling industry: as soon as the AI education system Call Annie was launched, the stock price of Chegg, an American education platform, plummeted by 50%, and AI teachers who do not need to make appointments, are currently free, and chat 24 hours a day may eliminate all language teachers, basic education teachers, translators, and even university teaching assistants in the future; The hair color can be adjusted at will, and the story of the AI model can be rewritten in the subsequent commercial operation, and the AI model Aitana Lopez is now completely able to undertake 10,000 euros of advertising per month.
As long as we are adaptable, the future of humanity may not be bleak. Science fiction writer Mu Ming said in an interview with Interface Culture that when machines can replace a lot of work, we need to examine whether we should pay attention to this kind of work. When machines quickly Xi learn the style of language, writers will figure out how to write more logical texts with more implicit content. As Edward Ashford Lee (Note: Plato and Technonerds, author of Co-Evolution: The Future of Human-Machine Convergence, a computer scientist) proposed, technology and humans evolve together, and humans will change themselves with the evolution of artificial intelligence. Mu Ming has a wild idea that in the future, ChatGPT can be used as a reference line or unit of measurement for innovation, and if a new novelist is launched, he can measure whether his creativity is 0.8 times or 1.5 times that of ChatGPT. If machines can learn Xi through humans, why can't humans learn Xi through machine feedback?
03 AI Risk and Regulation: How to Move Towards a "Human-Machine" Future?
Keywords: AI competition, AI illusion, AI war, AI legislation, Anthropocene
The societal risks posed by AI should not be underestimated. Even tech freaks like Elon Musk led the signing of an open letter in March this year, calling on all AI labs to immediately suspend training AI systems more powerful than GPT-4 for at least six months. However, some people believe that he is just delaying the development of AI products for his company.
In any case, Musk's call has not been answered, and the AI race is still in full swing. There are a number of points to be wary of in previous review articles by Interface Culture. For example, the U.S. Police Department uses predictive analytics software to patrol and search low-income neighborhoods more frequently based on its guidance, and an over-reliance on AI exacerbates pre-existing poverty and injustice. In addition, ChatGPT often "talks nonsense in a serious way" and gives wrong answers that seem to be very authoritative. People must be wary of this "illusion" and retain the Xi of self-verifying information.
British philosopher Elizabeth Miller also mentioned the relationship between artificial intelligence and war machines in an interview with Interface Culture. She noted that in the midst of the recent Israeli-Palestinian conflict, Israel is using the Habsora artificial intelligence system, known as the Mass Assassination Factory, which has killed thousands of people in the past few weeks.
At present, all countries are in full swing to promote AI legislation. In July, the mainland issued the Interim Measures for the Management of Generative AI Services. In October, Biden signed an executive order on AI. In December, the European Commission reached an agreement on the Artificial Intelligence Act, making it the world's first comprehensive regulation in the field of artificial intelligence, covering copyright, data privacy, public safety and other aspects.
In the future legislative direction, the democratic management and supervision of AI is crucial. As Kohei Saito, the author of "Anthropocene", said in an interview with Interface Culture, almost all the technologies we have now are funded and monopolized by private companies such as GAFA (Google, Apple, Facebook, Amazon), and ordinary people do not have the ability to develop technologies of this scale, even if a company induces them inside the black box, we have no way of knowing. So make rules to make sure it's not influenced by corporate logic, decide how to manage it in a democratic way, and have more oversight.
When tools become too powerful, everyone needs to think about how to interact with tools in the new era. In the article "The Emergence of the "Other": A Discussion on Generative Artificial Intelligence Large Models" in the surging thought market, Zhang Yuxin, a doctoral student at the Institute of Automation, Chinese Academy of Sciences, and Dong Weiming, a researcher at the Institute of Automation, Chinese Academy of Sciences, pointed out that when programmers no longer use 01 coding to communicate with computers, but use Python, a programming language that is closer to natural language, when people are learning Xi use prompt words every day to better communicate with AI, humans and machines seem to be merging with a language co-created by humans and machines. In this era of "human-machine integration", we must not only improve our technology, but also reflect on ourselves and deeply understand the limitations of human beings and the potential of artificial intelligence. Artificial intelligence will ultimately be relevant to each of us, whether at the level of creativity, work, or human self-awareness and understanding of language.
Resources:
Lee, Hye-Kyung. “Rethinking Creativity: Creative Industries, AI and Everyday Creativity.” Media, Culture & Society, vol. 44, no. 3, 2022, pp. 601–12,
https://authorsguild.org/news/thousands-sign-authors-guild-letter-calling-on-ai-industry-leaders-to-protect-writers/
https://authorsguild.org/news/ag-and-authors-file-class-action-suit-against-openai/