Nowadays, if you don’t write about ChatGPT, it’s like you’re not keeping up with trends and algorithms; if you don’t write about AI, you’re almost embarrassed to call yourself a tech writer. The last time I wrote about AI was in 2020, not counting the human-powered AI. I didn’t jump on the bandwagon not only because many people have already written about it and I don’t have enough deep understanding to offer more insights, but also because I care more about improving production relationships rather than increasing productivity.
Let’s start with a thought experiment: If you had to choose one of two paths to ensure that all humans have enough to eat and wear, would you choose to increase overall economic growth or reduce wealth inequality?
In 1985, Deng Xiaoping said that “some people should get rich first,” implying the choice of the first path. Thirty-six years later, in 2021, Xi Jinping proposed shared prosperity, emphasizing the second path. After the failure of international communist experiments, most people advocated for increasing overall economic growth and then taking care of the poor through taxation and social welfare, preferring the poor in developed societies over the middle class in a stagnant economy.
I don’t mean to say that the above thinking is wrong, but what I want to point out is that the logic is based on the premise that the global supply of food and other resources is slightly higher or lower than the overall demand. What if technology has advanced to the point where total productivity far exceeds the needs of the global population? Then we can reasonably infer that even if productivity no longer grows rapidly, as long as resources are distributed more fairly, or more precisely, more justly, it is already enough to support the world’s seven billion people.
Advanced countries have been discussing and even attempting to implement universal basic income (UBI). With AI developing rapidly, the day when global productivity far exceeds overall demand may not have arrived yet, but it is not far off. The key issue is, by then, how concentrated the world’s wealth will be and how widespread poverty will be.
AI boosts production efficiency
Throughout history, mastering the means of production has ensured an unbeatable position in the production relationships. The core means of production, besides the constant factor of land, have evolved from steam engines during the Industrial Revolution to oil, then modern information technology, and now data. AI is the fusion of information technology and data, using the former as the machine and the latter as fuel. As a result, data is often considered the oil of the AI era.
Although AI literally means artificial intelligence, it is more strictly defined as machine learning, which is a method that allows machines to discover underlying logic by processing large amounts of data.
As long as users clearly define their needs, programmers can develop software to solve the problem. Writing code, in essence, is an “IFTTT” (If This Then That) process. If a solution cannot be developed, it is often not due to the programmer’s technical abilities, but rather the user’s inability to clearly define their needs.
“Find the dogs in these 10,000 photos,” a request that a boss may think is very clear, is not at all clear to a computer because there is no definition of what a dog is. The boss may not be an ordinary person, but they are simply unable to clearly explain how to identify a dog; subordinates rely on their personal abilities to make judgments. This discernment is the “intelligence” of humans.
When we were young, we learned to distinguish between humans and animals and between cats and dogs. Although we were more or less taught by our parents and teachers, we had to use our own intelligence to find the common features of all cats and dogs, because their descriptions, no matter how clear, were far from sufficient for a non-intelligent computer to distinguish between cats and dogs using an “IFTTT” approach.
For tasks where the requirements cannot be clearly articulated, current technology can only handle them through machine learning. In the previous example, this would involve providing a large number of examples for the machine to learn how to identify a dog, which is the so-called “artificial intelligence.” The required materials are the dog photos we upload to Instagram every day, labeled with #dog, commented with “Cute dog 🐶,” and identified through reCAPTCHA.
The rapid development of AI in recent years is the result of big data combined with exponential improvements in chips. Only by mastering vast amounts of data and computing power can machines be trained to perform specific tasks, such as playing Go or the currently popular automated chatting.
Production Relations: Planting Melons Yields Melons, Assembling iPhones Yields Nothing
Production relations are a concept in Marxist theory. Marx believed that in order to live, humans must interact with society, producing, purchasing food, goods, and services. Production relations broadly refer to what tools an individual possesses, what role they play, and how much distribution they receive in the production process.
In layman’s terms, you work day and night without stopping, earning barely enough wages to make a living, while your supervisor directs you and earns twice as much as you do – this is production relations. The supervisor’s boss is the CEO, earning a hundred times your wage. When the company needs to cut costs, they lay off you and your supervisor instead of cutting their own salary, which drives up stock prices and earns them a hefty bonus – this is also production relations. The company’s original founder who came up with the idea and filed a patent is long retired and continues to earn even more than the CEO – this is yet another aspect of production relations.
Foxconn workers assemble iPhones day and night but do not own Apple stock, possess any technology patents, control iPhone production methods, or even afford to buy an iPhone. Planting melons yields melons, assembling iPhones does not yield iPhones – this is a typical representation of modern production relations.
Parents have always urged us to study hard and excel. Analyzing from the perspective of Marxist theory, this is because studying hard helps us master the means of production, occupy advantageous positions in production relations, and avoid being replaced by technology.
Today, we realize that our parents have led us astray. After twenty years of hard study, we find that we can simply press a button and have AI do our work. Knowing this, we might as well have become construction workers, who at least do not worry about being replaced by AI for the time being.
Blockchain Improves Production Relations
AI aims to enhance production efficiency, while blockchain focuses more on improving production relations.
In German, production relations are originally referred to as Produktionsverhältnisse. Verhältnis in this context can mean both relation and ratio or proportion, which coincides with the strengths of blockchain technology. When used in DAOs (Decentralized Autonomous Organizations), blockchain empowers stakeholders, allowing everyone to participate in the governance and distribution of benefits within a community.
We can try to imagine how blockchain could intervene in ChatGPT’s production relations. As the Chinese saying goes, “All articles in the world are one big copy.” Ancient people foresaw ChatGPT many years ago and summarized its basic principles in a few words, which reflects great wisdom. When we marvel at the wonderful articles written by ChatGPT, it is fed by all the articles each of us has written in the past.
Of course, even if you are Jin Yong, you cannot claim that ChatGPT’s stories are inherited from your works, because ChatGPT truly integrates the essence of various sources, and its learning materials far exceed what any individual has written in their lifetime. Since everyone’s proportion is too small and the process involves learning, digestion, and output, ChatGPT does not need to split the bill with anyone, regardless of whether the machine learning material comes from user data or public content.
However, even if a drop of water in the vast ocean has an infinitely small proportion, the ocean is ultimately made up of these droplets. Whether we are aware of it or not, we constantly train ChatGPT every day by typing on our keyboards and clicking our mice. If blockchain can enable us to own our works, help trace their origins, and handle extremely small “nano-transactions,” there is hope to record individuals’ stakes in this production tool. Thus, when ChatGPT generates huge revenues without a master (aka today), we will not play the role of farmers and workers who stop working in the AI era’s production relations.
This not only applies to ChatGPT but can also be applied to AI-generated graphics and any other machine learning model. Midjourney can instantly create artworks that might take an artist a week to complete because it references a vast number of historical images. Some artists despise tools like Midjourney and even initiate boycotts against their users, including myself, who occasionally uses them to create weekly report covers. On one hand, I can understand the pain of traditional artists; on the other hand, AI not only saves costs but also provides services that traditional artists cannot offer. In this regard, even if I am willing to bear the “AI royalties” behind machine learning works, current AI does not provide such options, but blockchain, which can locate origins, record ownership, and execute trivial transactions, has the potential to move in this direction.
The most in-depth discussion of similar ideas can be found in the book “Radical Markets,” which proposes “data as labor” and advocates for users to form “data guilds” to fight for rights and interests from AI companies. As the title suggests, it is a very radical approach.
99.99% of humanity will become the AI proletariat
As technology advances, ChatGPT and Midjourney are gaining popularity, while cryptocurrencies and NFTs are in a bear market. Some quick-witted netizens have made memes mocking blockchain technology. Bill Gates recently said, “AI is the big one. I don’t think Web3 was that big or that metaverse stuff alone was revolutionary but AI is quite revolutionary.”
If you say that the above example of how blockchain can improve production relations is not concrete enough, I would readily admit that while AI’s impact on work efficiency is increasing by tens, hundreds, or even thousands of times, blockchain’s role in improving production relations currently only has some basic concepts and directions.
However, this is not a reason for society to prioritize AI over blockchain; on the contrary, the significant improvement in production efficiency, which far exceeds the improvement in production relations, confirms the urgent need for society to devote more resources to the latter. This is also the underlying reason why I have always been fascinated by blockchain but rarely write about AI.
There is no doubt that AI greatly improves work efficiency, but if we cannot improve production relations in the AI era, we may face the most severe oligopoly in history, with most of the world’s population, including intellectuals, becoming the AI proletariat.
- Between justice and injustice, we read
- Artificial intelligence is not necessarily good, and human-centeredness is not necessarily right
My articles are not put behind a paywall and are all open for reading. If you like this article, please collect its Writing NFT to support writing and preserve journalism.