Phantom Magic: ChatGPT's Parasitism

　　Jos de Mul is an emeritus professor at the Department of Philosophy at Erasmus University Rotterdam, the Netherlands, former president of the International Society of Aesthetics, and a well-known scholar in the fields of philosophical anthropology and information philosophy, whose main works include The Odyssey of Cyberspace (Guangxi Normal University Press, 2007), The Tragedy of Finiteness: Dilthey's Interpretation of Life (Shanghai Joint Publishing Co., Ltd., 2013), and The Domestication of Fate: The Rebirth of Tragedy and the Spirit of Technology (Rotterdam, 2016). The Essence of Man-Made: The Road to Homo Sapiens 3.0 (Rotterdam, 2016), among others.

　　Chen Xin is a distinguished professor of history at Shanghai Normal University and a visiting professor at the Department of History at the University of Virginia.

　　ChatGPT is amazing! We love to revel in the possibilities of generative AI models, but shouldn't it be called "CheatGPT"? What would such a language robot be without humans?

　　Will generative AI become the golden grail of commerce?

　　The history of Artificial Intelligence (AI) and artificial neural networks can be traced back at least 75 years ago to the seminal work of British mathematician Alan Turing. Since then, the road to "artificial general intelligence" (i.e., machines that can perform any intellectual task like or even better than humans) has been paved with the promise of a reality within a decade. Many experts also see ChatGPT as a major breakthrough on the road to the Golden Holy Grail. And since the research institute OpenAI made version 3.5 of this artificial intelligence available for free in November 2022, the general public has also been fascinated by it.

　　Science fiction writer Arthur Clark believes that any sufficiently developed technology is no different from magic. If this statement applies to anything, then at this point ChatGPT couldn't be more appropriate. However, despite the large number of complex statistics involved, the principle is actually quite simple. Its full name is "Chat Generative Pre-trained Transformer", also known as "a generative AI based on pre-trained language models." It is a text generator that statistically calculates the most likely continuation when a string of words is entered, and then presents it to the user. For some time now, smartphones and word processors have known a simple version of this automatic replenishment. When you type the first letter of a word, you'll be prompted for the next word, so you don't have to type the whole word. In the Google search engine, predictive text has gone even further. If you type in "the capital of France," the search engine will show you a number of options, including "Paris is the capital of France."

　　In ChatGPT, the scale of autocomplete has increased considerably. The interface is in the form of a chat box, and you can give various commands to your virtual "conversation partner" with a flashing cursor. For example, you can ask it to translate or summarize a piece of text, make suggestions for the plot of a novel, movie, or video game, correct errors in a computer program, explain a scientific theory or political event, or give practical advice about health, gardening, or sustainable living.

　　ChatGPT's performance is often astounding, thanks to a huge language model created by OpenAI using a supercomputer with 285,000 processors (a number that is still increasing dramatically). In addition to the increase in computing power and the massive amount of data, this leap is also due to the launch of a neural network "transformer" in 2018, which uses an "attention mechanism" to focus on the most relevant words in the text. In addition, unlike older neural networks, it can process input data in parallel. Even so, these models often take months to develop enough parameters (the values that determine the conversion from input to output) to produce satisfactory results.

　　Since its inception at the end of 2022, ChatGPT has developed spectacularly. As of January 2024, OpenAI already has 180 million users and visits to OpenAI's website more than 1.6 billion times per month. Although OpenAI was originally a non-profit organization, users are required to pay $20 per month for version 4.0 as of March 2023. Then, in September 2023, "DALL-E" (a portmanteau of the painter Dali and the animated film Wall-E) was introduced, and in February 2024, the "Sora" video model was released, which generates ultra-realistic one-minute movie clips.

　　Needless to say, Open AI isn't the only company entering this market. In early 2023, Microsoft invested $10 billion in ChatGPT and integrated it into its own AI chatbox, Bing. Meta has launched LLAMA, Google has launched Gemini, and Apple is preparing to build the technology into its new operating system. Elon Musk and Jeff Bezos have also announced their own AI variants in 2023. Silicon Valley also apparently wants generative AI to be the golden holy grail of business. Development is moving at a rapid pace. In February 2024, the team of "Deep Mind" Demis Hassabis developed the Gemini 1.5 Pro, which allows users to enter up to a million characters. With it, the creators claim to achieve "long-context understanding". In March 2024, Musk open-sourced his AI system, Grok.

　　However, in addition to "content", generative AI has sparked a heated social debate about the usefulness and disadvantages of this form of AI. One camp sees AI as a panacea for many of the crises facing humanity, while the other, led by philosophers like Nick Bostrom and Max Tegmark, warns that AI excellence will surpass humans. Proponents of such utopian prophecies include many designers and funders of generative AI. For example, in March 2023, Demis Hassabis, Elon Musk, and Apple co-founder Steve Wozniak, among others, published an open letter calling for an immediate six-month moratorium on AI development in order to establish protocols and review systems to ensure AI security during this period. By the way, this pause did not materialize.

　　Symbiosis is a key factor in the evolutionary process

　　How should we understand generative AI and evaluate opposing values? I think the evolutionary and symbiotic perspective, and more specifically, the concept of techno-symbiosis, provides a good entry point into this.

　　The concept of symbiosis means "to live together" in ancient Greek, but in twentieth-century biology, this concept has not been fully studied. The neo-Darwinism that prevails among them holds that life is inextricably linked to the competition and struggle between selfish individuals, and that this is driven by selfish genes. It is no coincidence that American paleontologist Jay Gould believes that this view of humanity is very similar to neoliberalism. Gould argues that the economist Adam Smith, in his The Wealth of Nations (1776), not only explained the operation of the market in terms of the self-interest of individual actors, but also inspired Darwin to see self-interest as the driving force of the evolution of life in On the Origin of Species (1859). While Smith and Darwin also acknowledged the role of cooperation, later (neo)Darwinists focused almost exclusively on competition. The ideology of "greed is virtue" reached its zenith, fueled by biologists such as Richard Dawkins, author of The Selfish Gene (1976), and government leaders such as Margaret Thatcher (who advocated "no alternative") and Ronald Reagan (who joked about "nuclear bombing them").

　　While this ideology is still prevalent, the role of symbiosis in evolution and economics is now prominent. Regarding symbiosis, it can be divided into three basic types: symbiosis of the same species (animals that live in groups, in herds, and in groups), symbiosis of different species (such as flowers and insects), and symbiosis and abiotic symbiosis (from beaver dams to smartphones).

　　Symbiosis is a key factor in the evolutionary process for at least two reasons. First and foremost, it is an important force for innovation. Lichens are a good example. Lichens have long been thought to be separate biological species, but in 1877, it was discovered that lichens are a symbiont of fungi and green algae or cyanobacteria. These composite organisms emerged about 500,000 years ago, breaking down rocks and turning them into dirt, creating an entirely new ecosystem on a previously rocky Earth in which new land-based species could develop from aquatic organisms.

　　The endosymbiosis theory, proposed by microbiologist Lynn Margulis in the sixties and seventies, clearly shows how important symbionts are in terms of their ability to innovate. For a long time, neo-Darwinists ignored this, but now it is widely accepted. She points out that eukaryotic cells (cells with a nucleus and organelles) emerged about 1.6 billion to 1.8 billion years ago as a result of the symbiosis of two simpler single-celled organisms (bacteria or archaea), with one organism merging into the other, becoming an "energy factory." This new type of cell formed the basic unit of the entire animal and plant kingdom that would later be. In her book Symbiotic Planet (1998), Magris points out that symbiosis is the driving force of all evolution. (Produced by the "Thought Workshop" of the social science newspaper, the full text can be found in the social science newspaper and its official website)