The Colors of Superintelligence - Horizons of Knowledge Interwoven with Speed, Aggregation, and Quality (2025)

The Colors of Superintelligence - Horizons of Knowledge Interwoven with Speed, Aggregation, and Quality (2025)
The term "superintelligence" does not just refer to computers that compute fast or systems that can handle vast amounts of data. In Bostrom et al.'s discussion, it is envisioned as an entity that greatly surpasses humans in almost all intellectual domains, including scientific creativity, strategic thinking, and social intelligence. This is not "AI that is a little smarter than humans," but intelligence on a scale that would allow it to redesign the world while looking down on humanity as a whole.
To get a sense of what this might look like, Bostrom describes three aspects of superintelligence. The first is "speed-type superintelligence," an intelligence that has a human-like mode of thinking but whose speed is enhanced thousands or millions of times. A second to a human is perceived as several hours by this intelligence, and it is imagined as a being that writes a doctoral dissertation in a few minutes. The second is "collective superintelligence," a system in which a large number of intelligences are networked together to form a whole that is capable of solving problems better than any current organization or nation. This is similar to the image of the entire Internet acting like a single giant brain. The third category, "qualitative superintelligence," is described as an intelligence that thinks as fast as humans, but has a fundamentally different depth of understanding and insight structure, and can freely handle concepts and strategies that humans cannot grasp.
What these classifications show is that the word "smart" encompasses a wide variety of dimensions. If it were merely fast processing, it could be understood as a simple scaled-up extension of human beings, but if we include collectivity and qualitative leaps, we open up areas that are beyond the reach of human intuition. For example, qualitative superintelligence is portrayed as an entity that can replicate the difference in intelligence between humans and animals between humans and machines. The suggestion is that just as a squirrel cannot understand quantum mechanics, there may be modes of thought that humans can never attain.
On the other hand, contemporary research and commentary has also begun to view humans themselves as the "superintelligence we are looking for. By redesigning advanced network and quantum technologies as an infrastructure open to society as a whole, rather than for a few powerful people, the "collective human intelligence" could function as the original superintelligence. This idea does not project superintelligence onto machines alone, but rather redefines the structure of human society as a whole as another form of intelligence.
Thus, superintelligence is not a fixed concept that can be divided into a single definition, but an entity that can be depicted three-dimensionally on multiple axes, such as speed, collectivity, quality, and relationship with human society. Depending on how the contours are sliced, both the risk assessment and the expected benefits can vary greatly. The diversity of definitions itself is an unavoidable point of contention when considering how to deal with the intelligence of the future.