[Humanity]
Regardless of whether this world originates from a “Creator” or from what Laozi referred to as “nature,” humanity undoubtedly exists with a special significance, largely because we naturally stand from a first-person perspective of anthropocentrism. Until extraterrestrial civilizations are proven to exist, humanity will continue to occupy the throne of the highest known intelligent life form. Currently, research and development in fields related to AI, superintelligence, and robots are in full swing, with many discussions and concerns similar to those about “AI/robots replacing humans” and “carbon-based civilization vs. silicon-based civilization,” which cannot be dismissed as entirely unfounded. Therefore, as both the initiators and the challengers, it is indeed necessary for humanity to conduct a thorough examination and comparison of itself and its potential challengers, both to better promote the development of AI/robots and perhaps to find a future “Shangri-La” or “Zion” that can provide the ultimate refuge for humanity.
One can draw a parallel: the relationship between the “Creator” and humanity is akin to the current relationship between humanity and AI/robots, where humanity is both a creation of the “Creator” and plays the role of the creator of AI/robots. In the foreseeable future, the “Creator” should be far superior to humanity, thus the works of the “Creator” (i.e., humanity itself) are indeed worth studying to provide reference for the research of AI/robots.
Previously, in Reflections (53): Thoughts on the “AI Data Center Construction Bubble” News, I raised some questions comparing human intelligence and artificial intelligence. Understanding and answering these questions will greatly influence our path and ultimate form in achieving General Artificial Intelligence (AGI). Whether in social sciences or natural sciences, great innovative practices are inseparable from correct theoretical guidance. It is somewhat like the current dilemma faced by particle physics, where, in the absence of significant breakthroughs in fundamental theory, many hope to build a super collider to try their luck. Currently, the large models of AI based on ultra-large-scale neural networks are racing ahead in scaling up, increasing computing power, and optimizing algorithms, yet they seem to have yet to truly unlock the black box of intelligence to understand both its existence and its reasons. Thus, where the path of large models will ultimately lead, and whether it can achieve true AGI, remains a significant unknown.
As a highly intelligent life form, is there an inevitable connection between “intelligence” and “life”? That is, is the level of intelligence and the movement of life a dialectical unity that cannot be separated (with existence and non-existence mutually arising, and being interdependent)?
From the smallest individual to viewing all of human civilization as a whole, our path of life and intelligence development evolution generally includes the following key elements:
——Experience/Experience: This involves being in various different scenarios, experiencing countless people, objects, and events;
——Perception: This includes the five senses brought by physiological organs (sight, hearing, smell, taste, touch), as well as the psychological awareness level of “emotion” (for example, in Chinese, terms like “joy, anger, sorrow, and happiness,” “state of mind,” “empathy,” etc., and in English, “I can feel you”), along with more mysterious intuitions, premonitions, sixth sense, seventh sense… It can be said that humanity is a highly complex and sensitive multifunctional sensor complex.
This may be a unique advantage we have over AI/robots. In terms of a single sensory function, the skin of a human being may already be a “god-level equipment” that robots can only aspire to, not to mention the more advanced and mysterious “emotions,” sixth sense, and seventh sense. The greatness of combining so many different sensory functions in collaboration can be slightly understood by considering the debate in intelligent driving between “pure visual routes vs. LiDAR-assisted routes.” The disadvantages in perception, while the mechanisms affecting intelligence levels are not yet fully understood, are likely to cause certain deficiencies or defects, which can be somewhat evidenced by observing individuals within the human population who have sensory functions that are impaired or absent.
——Learning & Cognition: This refers to the series of activities of thought and consciousness we engage in regarding the “world model” formed by the aforementioned experiences and perceptions projected in the “intelligent system” of the human body (perhaps not limited to the brain), including symbolization, reasoning (induction, deduction, abduction), memory, etc., encompassing both metaphysical and physical aspects.
Unlike the neural network large models based on Bayesian theorem/probability theory, this human learning and cognition system seems to be more established on a symbolic logic paradigm. Humanity’s abilities in perception and understanding, logical reasoning, learning new things, exploring the unknown, drawing inferences, and cross-domain applications currently appear to be fundamentally superior to those of large model AI. In terms of efficiency, although humanity is at a disadvantage in overall computing power, when considering the energy consumption and output efficiency per unit of computing power, human intelligence may actually be the better option for the world.
——Interaction: Through interactions with the external world, we not only enrich our life experiences but also apply and train our intelligent systems. In this process, we continuously “strengthen our bodies and civilize our spirits,” leading life forms and intelligence levels to gradually mature.
——Organization: This refers to the organizational forms of human groups themselves. From small families or groups to entire nations and even the world, from the earliest primitive groups to modern international societies, humanity has always existed in some specific organizational form (rather than as individuals). Although human history is filled with various conflicts and disputes, and organizational forms vary in different periods and contexts, there has always been a pursuit of the ultimate goal of “harmony in diversity” (ideal states, world harmony, and a community of shared human destiny). This flexibility, adaptability, and ultimate pursuit embody the strong vitality of humanity as a species and civilization.
Corresponding to AI/robots, it is still uncertain what organizational forms they would adopt if AI/robot civilizations were to emerge in the future. In many science fiction films and works, a common structure seems to be a centralized form similar to a swarm: a few super “leaders” acting as queens, while the rest are numerous, clearly defined “worker bees” (Agent). This structure is undoubtedly the most efficient in terms of organization, naturally aligning with the value standards of AI (seeking optimal solutions for efficiency functions), but it also has significant flaws.
——Invention: This includes all human creations on both material and spiritual levels, which is key to distinguishing us from other life forms on Earth and forming “civilization.” From the logical chain of “‘Creator’ – World – Humanity,” it is through invention and creation that humanity began to feedback to this world and even to the “Creator,” achieving a certain closed loop: that is, in addition to experiencing, perceiving, and understanding the world, we further expanded the existence and connotation of the world, perhaps even making “God smile.”
In addition to the above elements, human civilization possesses some characteristics that AI/robots may not have.
——Vitality: Life, aging, illness, and death, renewal. From a physiological perspective, genetic recombination and mutation ensure the renewal vitality of human life forms. From a moral and ethical standpoint, Confucius said, “If one does not understand life, how can one understand death?” Conversely, it is also valid: without understanding death, how can there be life? The understanding and attitude towards life and death have always been an important component of various cultures among different human ethnicities. The extent to which AI/robots can truly understand this will greatly influence their self-awareness.
——Inheritance rather than replication: Because individual life is limited, human civilization must be passed down through generations relying on various carriers. However, later generations cannot simply copy and paste; they must internalize through learning, inevitably encountering confusion, questioning, and even overturning, thereby effectively promoting the overall progress of civilization. This model of “inheritance – questioning – overturning – progress” is precisely the opposite of the overfitting phenomenon present in large model AI.
——Diversity and inclusiveness nurture genius: From the perspective of culture and civilization, there is no need to elaborate. Just from a physiological standpoint, genetic mutations fundamentally ensure that life forms can maintain diversity over the long term. While most of these rare mutated individuals may be disadvantageous, they may also give rise to several great geniuses among millions, leading humanity forward. In contrast, the probability-based large models of AI fundamentally stifle the possibility of such “mutations”: the model results that pursue maximum likelihood probabilities will prioritize the exclusion of long-tail low-probability events. It is akin to Agent Smith in “The Matrix,” who is exceptionally capable but is considered a bug in the Matrix system that must be eliminated.
——The pursuit of the “other shore”: Humanity’s exploration and pursuit of the ultimate mysteries of the world and life are not limited to the realm of religion; they are deeply rooted in the core of human civilization and exist, to some extent, in the depths of each individual’s heart, whether overtly or covertly. The “other shore” determines people’s worldview, values, and outlook on life, although it may exist in different forms and connotations among different human groups. Correspondingly, how to instill “correct” values in AI remains a scientific challenge.
……
Unknowingly, I have written these words, yet they seem to serve no purpose, merely a thought exercise. Too much of anything is as bad as too little; I shall stop here.
* * *