In 2018, David Graeber's Bullshit Jobs: A Theory was a hot topic when it was published. Recently, with the release of ChatGPT, OpenAI's blockbuster report showed that it could affect 80% of jobs, once again pushing the discussion about "shit work" to the forefront.
Many people worry that AI will take our "shit jobs" and eventually replace humans. Whose job will AI take first? When will we achieve a 15-hour workweek? How can we not be eliminated by the AI era?
Knowledge Vientiane invited Liu Yongmou, professor of the School of Philosophy of Chinese Minmin University, economist Liang Jie, doctor of the Institute of Computing Technology of the Chinese Academy of Sciences, Niu Xuesong, and former teacher of the University of Science and Technology Beijing, Ji Yanjiang, to explain AI and your "shit" work in detail.
Core points
- 1 Jobs that do not deal directly with people are more likely to be "snatched" by AI, and the existing social value ranking will face impact.
- 2 The role of AI in reducing human labor time is not only a matter of technological development, but also affected by social systems.
- 3 To judge whether AI is self-aware, the main basis is whether it has subjective behavior, whether it will automatically do logical reasoning, or whether it will have partial personalized behavior.
- 4GPT puts many people on the same line, and the competition is no longer the ability cultivated by the traditional teaching system, and the competitors are no longer just humans.
1. How powerful is GPT? Will the ranking of human social values face a reshuffle?
Ji Yanjiang: First of all, please introduce your experience of using ChatGPT or GPT 4.0.
Liu Yongmou: GPT performs differently in different fields, and its iteration is also very fast with technological development and data updates. Some people say that GPT often "seriously", but according to professional linguistic research, people often talk nonsense. As a set computer program or application, it can get rid of presets and nonsense, and that level is also very high, I don't think this is a bug or a problem.
Xuesong Niu: I mainly use it as a retrieval tool, using it in more deterministic situations, and rarely have chat-style communication with it. In the past, we needed to have a strong ability to identify information to apply the search results of the search engine to work, and GPT can directly give you a piece of code or information that can be used directly according to your needs, which greatly improves work efficiency.
Liang Jie: I started using GPT at the end of last year, but I always feel that I am still "knowing" it. At the beginning of the year, it was able to reach the level of high school students and even college students, but it could not give in-depth answers in the field of expertise, and occasionally even made up wrong answers and false references. However, with the introduction of GPT-4, people have become more proficient in asking and using it, and many of the problems of the past are being corrected.
In my daily work, drafting documents and outlines, and writing simple code can be handed over to GPT. However, it seems that there is still a certain gap in requiring it to carry out in-depth academic research assistance, and it may be possible in the near future.
Ji Yanjiang: The magic box of AI is being opened, and recently there is Auto-GPT, which can break down a task into several steps to automatically execute, such as automatic Internet search, shopping, and email, further reducing manual intervention. This shows that GPT-based applications have great potential in the future.
Recently, BlueFocus has announced the elimination of copywriting outsourcing, AI seems to be constantly impacting the job market, which industries and practitioners will be the first to hit? Will it change the social value ranking?
BlueFocus announced the complete suspension of copywriting outsourcing (Source: The Paper)
Liang Jie: The replacement of human resources by AI is very worth discussing. In the past, we sorted different labors according to the degree of human capital accumulation from low to high, for example, restaurant dishwashers do not require academic qualifications, and programmers may need years of education and require human capital accumulation. Therefore, we used to think that AI or robots should replace our labor in this order, from lowest to highest, that is, Marx said that complex labor is more difficult to replace.
But after the advent of AI, the first to replace programmers or translation industry practitioners who need years of learning are the same. Although the work of hospital nurses, which is closely linked to human subjective consciousness (which requires observing the patient's physical condition), does not require much learning, it is less affected by the standard process of the machine, and the impact is much different from our assumption, which is the most direct challenge.
In turn, as AI develops, a large number of new jobs will be created to "serve" AI and assist it to perform more capabilities. So employees who seem to be directly replaced by AI in the short term actually have the opportunity to enter new areas of work.
In general, the industries most affected by AI are those that do not require direct contact and communication with people at work.
2. How to achieve a 15-hour work week in the AI era? In addition to technology, social institutions need to be considered
Ji Yanjiang: Many people joke that in the past, the job was to equip people with a computer for the company, but now the job is to equip a person for the company's computer. There is currently a narrative that AI will lead to so-called "Industry 4.0" that greatly improves social efficiency, which means that the same number of people need less time to create the same amount of social value.
The economist Keynes predicted in 1930 that humans might not work more than 15 hours a week. Do you think that with the blessing of AI in the era of Industry 4.0, can we achieve a working week of no more than 15 hours?
Liu Yongmou: The famous American anthropologist Graeber once put forward the idea of technological slowdown, which is different from what we usually think that the faster the development of technology and the faster the pace of accelerating social development, he believes that the real acceleration technology is the technology of monitoring and controlling others.
David Graeber is a famous American anthropologist and author of books such as "Bullshit Jobs". (Image source network)
Many scholars have measured the question: when and how many products are needed to make all human beings comfortable? In 1929, the technocracy Scott and others measured from the perspective of converting matter into energy, and measured according to the 25~45-year-old people working 6 hours a day, working four days a week, and working for 20 years in North America, and concluded that if the goods and services produced in North America were evenly distributed, they could already meet the comfortable life of all people.
Galbraith, a scholar of institutional economics, proposed the "rich society theory", identifying the above point in time as the sixties and seventies of the last century. It is believed that the large amount of materials brought about by the development of social productive forces is enough for us to live a prosperous life, and that social poverty is only relative poverty rather than absolute poverty, shifting from the problem of production to the problem of distribution. For example, 60% of the world's jeans are produced by Chinese, but not every Chinese has jeans to wear.
Institutional economist Galbraith (source network)
The above examples fully show that the horizontal between the development of AI technology and our working hours is not entirely a technical problem, but more a social system design problem. If the system does not change, it is difficult to say that our working hours will decrease. Of course, we can also see that the general trend is still technological production to shorten the total human working time: in the early days of the Industrial Revolution, workers worked 16 hours a day, and after the labor movement we had weekends and eight-hour workdays.
Returning to the issue of AI technology, most scholars and experts believe that from the perspective of the future, the vast majority of human physical labor will be replaced by robots, which I call "robot labor society". The word robot first appeared in the Czech word for machine labor, and why invent it if you don't let robots work for us?
From a near-term perspective, this substitution has a gradual process, accompanied by changes in the social system. In terms of skills, human cognitive abilities have various aspects such as observation, expression, reflection, criticism and memory, while GPT is mainly memory, organizing information and simple reasoning. Regardless of the industry, the conflict between human skills and AI skills may lead to obsolescence. One way to solve this is to ensure that we are better at overlapping parts than AI, and one is to tap into our own creative and critical thinking capabilities that robots do not have, both of which may help us better adapt to the future.
All in all, the relationship between AI and human resources is a combination of new technological development and institutional change, and cannot be predicted purely from a technical perspective.
3.AI Does it develop self-awareness? What are the odds of superintelligence replacing humans?
Ji Yanjiang: Some scholars worry that human beings will eventually be replaced by artificial intelligence, which involves the problem of "self-awareness", because only after artificial intelligence awakens itself will it have subjectivity, and it is possible to replace human beings as new intelligence in the universe. From a professional computer science or artificial intelligence perspective, is there a definition of machine consciousness now?
Xuesong Niu: There is no clear definition to judge whether an AI is conscious. Early judgment of whether AI is conscious is through the Turing test, that is, a person and a machine are placed in a black box, and the same questions are asked at the same time, and it is judged whether it is a human or a machine according to the answers of the two, and if the machine is judged to be a human, it successfully passes the Turing test.
Although many machines in some use cases can basically pass the Turing test, in my opinion, this is just a growing technology, not yet at the level of consciousness.
Turing test diagram (image source network)
Consciousness is actually like the ability to take the initiative, such as students will take the initiative to say hello when they see the teacher, and GPT will not take the initiative to greet you. So now to judge whether AI is self-aware, it is mainly from the perspective of whether there is subjective behavior, whether it will automatically do logical reasoning or whether there will be partial personalized behavior.
Up to now, the existing technology is still a partial statistical model, no emotional and subjective consciousness occurs, can not take the initiative to do some things, can not have a clear cognition of the surrounding environment and do some reasoning or behavior based on cognition. But maybe AI will actually become conscious one day in the future.
Ji Yanjiang: In the past, technology was more of a tool attribute, and now artificial intelligence is more like a small partner. Is it possible for AI to become a superintelligence that will replace us in the future?
Liu Yongmou: Let me mainly make four points.
First, since the birth of artificial intelligence, similar issues in the industry have been discussed many times. Every new AI breakthrough leads to a wave of discussion about the machine's self-awareness: Does AI have consciousness? When exactly is it conscious? This is also a kind of "hype", so as to gain a lot of money and social attention, that is, effective "artificial intelligence propaganda", but also make machines a popular science fiction theme (called "robot culture").
Second, what exactly is consciousness? Philosophy is the discussion of anthropology, the most important of which is the problem of consciousness in epistemology. I can tell you unequivocally that no one has figured out what consciousness is so far, whether human consciousness is in the brain or in the body, embodied? The jury is out. Human consciousness has not yet been clarified, let alone robot consciousness?
I prefer the behaviorist approach, where metaphysical words such as soul and consciousness should be eliminated from science. That is, we look at an organism, just give a stimulus and observe the feedback, just like the function, y=F(x), only discuss the relationship between x (stimulus) and y (response quantity), as for whether the intermediate process has a soul or consciousness, can be ignored.
Third, about the Turing test. Some believe that its birth is related to Turing's homosexual experience, the father of artificial intelligence, and he was inspired by issues such as how to play or be a woman as a man, and how homosexuality plays as a non-homosexual. The Turing test is actually a cultural phenomenon, and it also derives the gender Turing test.
Fourth, will robots rule humans when they are conscious? Many people have envisaged many avenues. For example, superintelligence has appeared, and it has left the earth directly without competing with humans for resources. But there are also robots that only like to make paper clips that regard humans as raw materials, kill all humans, and make paper clips from the microscopic particles that make up humans.
But I think at the moment it seems to be alarmist. Many people believe that the biggest threat now is climate change, and the French philosopher Latour said when we discussed the relationship between epidemic response and technological governance that the problem of deadly viruses is not as serious as climate change, let alone super artificial intelligence. If we don't address the higher-priority existential risks, we'll be extinct before super-AI emerges.
Ji Yanjiang: According to the Turing test, judging whether a machine is conscious may have certain limitations. As for the problem of understanding consciousness, I think the narrative about consciousness has been concretized after the introduction of GPT, it may be a neural network driven, we can train it with language, vision, touch and other information. Regarding consciousness, maybe we don't have a big breakthrough at the philosophical level, but when we talk about this open topic, our narrative becomes more nuanced, which is also a great improvement. The topic of AI replacing humans is indeed hype on the one hand, attracting more people's attention; On the other hand, it will also stimulate us to think seriously about social progress, national governance and other issues, and may have some inspiration for future social development.
4. How can we not be eliminated by the AI era? How are new technologies impacting existing education models?
Ji Yanjiang: In fact, every type of work has its "lifespan", and the rapid development of technology has brought about "no insurance" for young people to work and study. In the GPT era, what psychological cognition, knowledge reserve or ability improvement should ordinary students prepare for employment? How do you keep yourself from being left behind by the technological age after coming out of the university classroom?
Liang Jie: David Graeber's book "Shit Work" has been very popular in recent years, and people who like this book should be very happy to see GPT appear. GPT has eliminated a lot of bullshit work, and it's great that people put their energy into more creative and meaningful work.
In fact, our education system and knowledge system must be far behind the development of AI, and the update of the knowledge system handed down from generation to generation can only be promoted little by little, in decades, hundreds or even thousands of years; And the update speed of AI is not measured in years, months or even days. So TK teaching systems may be superior in personality formation and social interaction, but lag behind in equipping students with job-ready skills.
Bullshit Jobs: A Theory is a 2018 book by David Graeber that mentions the existence of meaningless jobs in society and analyzes their social harm. (Source: Douban)
AI development has placed entirely new demands on us, such as prompting computer engineers that training that has never been done before has become a very critical capability. GPT puts many people on the same line, and the competition is no longer the ability cultivated by the traditional teaching system, and the competitors are no longer just humans.
So our first step to learn and exercise is how to make better use of AI, such as breaking down the problem to be solved into different overall frameworks, which are completed by GPT and finally integrated. The second step is to further think about which jobs are not to be replaced in competition with AI, but this is difficult to accurately judge. For example, if AI generates a movie, are you willing to watch it? Robots play basketball faster and higher than people, and they can sing better than people in concerts, are you willing to watch it? GPT or AI is no longer a question of whether we learn or not, but how we can better coexist with it and make it from a mere tool to a better servicer.
Robot shooting at the Tokyo Olympics (Source: Xinhua)
This is a big challenge for our university students. For years, we have been brushing up on questions and exams, but we rarely think about how society should deal with itself when it is in the midst of a great technological change.
Xuesong Niu: Technology development is now more exponential, and I think as young people we still have to embrace change, learn technology and recognize its upper limits as much as possible, use this tool to improve work efficiency, and arm yourself with AI to improve your personal capabilities. In the future, the technological productivity represented by AI will far exceed that of humans, and we are likely to need to serve AI and make better use of it to bring more convenience.
Liu Yongmou: Today's young people may not be very interested in expert advice. The question of how students can ensure that they will not be eliminated in the AI era obviously cannot be pushed to students alone, telling them that "you should work harder, learn more, should run faster and jump higher". This problem involves the entire system engineering of society, and the consequences of change, whether good or bad, cannot be borne entirely by individuals. Therefore, in addition to giving advice to students, we should also think about how we can give students a better living environment to live, live, work and study in the future.
After the launch of ChatGPT, the country and the mainstream of society are most concerned about three issues, one is unemployment, the other is education, and the third is security including ideology. For schools, perhaps the first thing is to think about the future, and insist that the past education method is to carve the boat and seek the sword. In the future AI-assisted survival society, AI is needed to help or even replace work and learning, and students who can survive in the future environment must be cultivated. If you think about the problem purely from the perspective of skills or jobs, you can't learn a certain job for ten years, and you end up working for one or two years, which can only be exhausting.
In fact, we should think about whether skills overlap with robots, or whether the overlap goes beyond robots. For example, critical reflection ability and communication ability, the robot will only answer and will not ask questions, it does not have this ability. Another example is translation, although many intelligence has good translation ability, but if the level of translation is very high, it will not be eliminated by robots, it is not only a matter of translation, but also the ability of poets and literary literacy.
I teach at Chinese Minmin University, and ChatGPT has a great impact on the humanities and social sciences such as writing, translation, and finance, so I personally feel that the cultivation of liberal arts talents needs to have major adjustments, such as taking the elite route, and cultivating students to a high level, otherwise it is difficult to compete with robots. In the past, the liberal arts were erudite and poetry books were written, but now we can completely obtain them through instant search, so we need to move from erudition to reflection.
High-quality liberal arts education is actually quality education, giving us the ability to communicate and communicate that robots do not have, and life must have a sense of meaning and transcendence. For example, now that the liberal arts are classified very carefully, should we move towards a kind of "general learning" in the future, cultivate more critical thinking, problem-oriented, and creative and innovative abilities.
Guest introduction
- 1 Liu Yongmou Professor of the School of Philosophy, Chinese Minmin University, researcher of the National Institute of Development and Strategy
- 2 Liang Jie is an economist who teaches at Shanghai University of Finance and Economics
- 3 Xuesong Niu, Ph.D., Institute of Computing Technology, Chinese Academy of Sciences
- 4 Ji Yanjiang Roundtable host and former teacher of University of Science and Technology Beijing
Finishing: Wang Jiaxin
Copyright statement: This article is an exclusive contribution of Tencent News "Knowledge Vientiane", unauthorized media reproduction is prohibited, but welcome to forward to personal circle of friends.