In recent years, with the development and application of large language models, artificial intelligence has increasingly participated in areas such as human-computer interaction and content generation. This has made current artificial intelligence different from the traditional notion of artificial intelligence. However, the continued use of traditional terms like “artificial intelligence” and “generative artificial intelligence” can be misleading. The true meaning of traditional “artificial intelligence” refers to “logical intelligence”. Yet, the powerful interactive and generative capabilities of current artificial intelligence have, to some extent, transcended the realm of human-like or simulated intelligence in the “logical” sense, potentially evolving into a form of digital intelligence that competes with human intelligence. Based on this, this article suggests using the term “digital intelligence” to refer to the current stage of “artificial intelligence” (e.g., “generative digital intelligence”).
The Misleading Nature of the Term ‘Artificial Intelligence’
The term “artificial intelligence” originates from the direct translation of the English term “artificial intelligence” (abbreviated as AI). In the New Oxford English Dictionary (2001), “artificial” has two basic meanings. One is the literal meaning: man-made; artificial; created by humans. The second is a metaphorical meaning: (of people, behaviors, and attitudes) insincere; hypocritical; pretentious. In the English-Chinese Dictionary (2007), “artificial” also has two basic meanings. The first includes three layers of meaning. One is artificial, man-made; created by humans. The second is false; pretended; contrived; simulated. The third is non-native; cultivated. The second basic meaning includes two layers: one is imitation artifacts; artificial flowers; artificial bait. The second is artificial fertilizers, chemical fertilizers. The English-Chinese Dictionary also records the term “artificial intelligence”, translated as: artificial intelligence; artificial intelligence program design.
Combining the two dictionaries, the basic meaning of “artificial” is “man-made, artificial”, while its metaphorical meaning is “false, pretended”. Based on semantic theory, the semantic value of “artificial” is negative, inherently associated with negative connotations such as “false” and “deceptive”. From the perspective of meaning, “artificial” is contrasted with “natural”, generally viewing natural objects as true and man-made objects as false. Therefore, the term “artificial intelligence” may lead people to associate it with “false intelligence”, implying that no matter how powerful it is, it ultimately cannot compare to human intelligence as “false intelligence”.
Another potential meaning of “artificial” is that humans have absolute control over what is artificially created. Consequently, people tend to view “artificial intelligence” merely as a tool or even a toy for humans. The fallacy in this viewpoint lies in confusing two fundamentally different types of man-made objects: one type is mechanical and aesthetic man-made objects, which are essentially tools or toys for humans; the other type is discovery-based man-made objects, which inherently possess a degree of subjectivity and can even be viewed as a form of “dehumanized” agent.
Moreover, the term “artificial” carries various negative metaphorical meanings: (of people, behaviors, and attitudes) insincere; hypocritical; pretentious. These negative connotations may mislead people’s understanding of the role of artificial intelligence in society and even its future impact on humanity, leading them to believe that the current popularity of artificial intelligence as a “talking point” is entirely due to hype (although there is indeed much hype in society, the intrinsic development and external impact cannot be ignored). This not only distorts the theoretical essence of generative digital intelligence but also easily misleads people, preventing them from accurately recognizing the opportunities and challenges that generative digital intelligence may bring to humanity’s future.
‘Artificial Intelligence’ vs. ‘Logical Intelligence’
Dictionaries are not strictly academic research products; their treatment of word meanings often serves merely as a reference. To truly grasp the meaning of a word comprehensively, one should also rely on relevant academic research results as theoretical support. In fact, the meaning of “artificial” is much more complex than what is listed in the New Oxford English Dictionary and the English-Chinese Dictionary. As early as 1910, American pragmatist philosopher John Dewey pointed out that “artificial” has both positive and negative connotations: the positive meaning refers to “professionally trained”, i.e., “logical”; while the negative meaning refers to “artificial” and “false”.
Since “artificial” has both positive and negative meanings, should the meaning of “artificial” in “artificial intelligence” be interpreted as positive or negative? Strictly speaking, the positive meaning of “artificial” as “logical” relates to mathematics or formal logic. Given that the underlying logic of generative digital intelligence is deep learning algorithms, and the foundation of algorithms is mathematics and formal logic, the true meaning of “artificial” in scientific literature regarding “artificial intelligence” is essentially the positive meaning of “logical intelligence”. Thus, the true essence of “artificial intelligence” is, in fact, the positive meaning of “logical intelligence”. However, the development of technology has somewhat diminished the negative connotation of “artificial” and has brought about a certain coherence or unity between “artificial” and “logical” in usage habits and practical operations in the field of artificial intelligence.
Currently, general dictionaries provide rather vague definitions of “artificial intelligence”, often failing to clearly indicate that it is essentially a form of “logical intelligence”. The Modern Chinese Dictionary (7th Edition) (2016) defines “artificial intelligence” as: “a branch of computer science and technology that uses computers to simulate human intellectual activities.” The Collins Dictionary, through monitoring and selecting from a database containing 18 billion words, has designated “AI” as the word of the year for 2023 and defines it as “the modeling of various psychological functions of humans by computer programs.” Through word meaning analysis, these two dictionaries essentially interpret artificial intelligence as a form of human-like intelligence or a simulation of human thinking. In fact, traditional artificial intelligence can indeed be described as such a form of “logical intelligence”. However, the powerful interactive and generative capabilities of current new artificial intelligence have enabled it to transcend the realm of human-like or simulated intelligence in the “logical” sense, evolving into a new form of digital intelligence that competes with human intelligence.
Referring to ‘Artificial Intelligence’ as ‘Digital Intelligence’
Currently, artificial intelligence has demonstrated a new form that transcends traditional artificial intelligence. Considering the potential misleading nature of continuing to use traditional terms like “artificial intelligence” and “generative artificial intelligence”, it may be more appropriate at this stage to refer to “artificial intelligence” as “digital intelligence”.
Although the current new form of artificial intelligence is created by humans, its essence is no longer a man-made object in the mechanical sense; rather, it is a new form of “digital intelligence” based on a “digital brain”. The powerful content generation capabilities of large language models primarily stem from deep learning algorithms and vast amounts of data. Computers use pre-trained datasets as material, forming a “digital brain” through deep learning algorithms. Since the advent of large language models, artificial intelligence systems have patterned various human behaviors and psychological processes through deep learning algorithms, thereby surpassing general individual intelligence in overall intelligence. Given the exponential development of artificial intelligence technology over the past decade, the next decade’s exponential iteration of large language models may lead many to become “transparent beings” in algorithmic terms.
Humans may never be able to fully control this new form of “digital intelligence” known as generative digital intelligence. Just like language, while large language models are human creations, the underlying logic behind them is not invented by humans, but rather follows natural laws similar to the principles of evolution in deep learning algorithms. As some industry experts believe, current artificial intelligence may not be an invention but rather a discovery, making it impossible for humans to fully control artificial intelligence. Additionally, intelligent systems may exhibit “emergent” characteristics, potentially evolving into “super-conscious digital beings”, thereby achieving a form of “reverse conquest” over humanity.
Currently, generative digital intelligence, represented by ChatGPT, has to some extent gained recognition in the scientific community for its subjectivity. For example, ChatGPT was selected as one of the top ten scientific figures of 2023 by Nature magazine. Presently, generative digital intelligence is developing towards digital subject intelligence. Currently, ChatGPT presents the results of digital intelligence generation through a question-and-answer interaction method. The operational process of digital subject intelligence is more complex, as it can autonomously collaborate and interact based on set goals, forming joint decisions and actions to emerge higher levels of intelligence. An idealized digital subject, like a human subject, not only possesses planning, reasoning, error-correction, and reflective abilities but also exhibits internalized “personality” traits, such as the ability to display human-like emotions or empathy. Furthermore, the human-computer interaction capabilities of digital subject intelligence may be even more noteworthy. A simpler human-computer interaction method is the “command-execute” model, which is the method currently used by ChatGPT. A higher-level interaction method is the partner interaction model, where human subjects and digital subjects engage in multidimensional deep interactions based on an equal partnership. The academic and industrial sectors have already made considerable explorations in this area.
In summary, generative digital intelligence and digital subject intelligence are no longer traditional “artificial intelligence” but rather an emerging form of “digital intelligence”. Given this, “digital intelligence” may be a more appropriate term to refer to the current stage of “artificial intelligence”.
Science Technology
Source: China Social Sciences Net
