Can Generative AI Philosophize?

[PDF] Sakiya Arakawa, 'Can Generative AI Philosophize?', researchmap.

Introduction

     I wrote an article titled "Generative AI and Hegel". At that time, I raised such questions as "Will generative AI develop Hegel studies?", "Can humans understand Hegel as generative AI?", "Can generative AI transcend Hegel?" These questions are not limited to Hegel, but are interchangeable with other philosophers, whether Socrates, Hobbes, or Nietzsche. Since they are generally called "philosopher," the more fundamental questions are "Will generative AI develop the studies of philosophers?", "Can human understand philosopher as generative AI?", "Can generative AI transcend philosophers?" Then, the relationship between generative AI and philosophy becomes an issue. Already during the second AI boom, issues such as the Chinese room (Searle, 1980) and the frame problem (Dennett, 1984) were raised. However, the status quo of generative AI is changing, and in particular the advent of neural networks has brought about significant changes in AI. Incorporating this singular change, the relationship between generative AI and philosophy must be questioned anew.

Can Philosophy Be Translated into Vectors?

     I have a hypothesis: "Can philosophy be translated into vectors?" I am not certain about this hypothesis. I would like to test whether this hypothesis is correct or not. The reason why I have such a hypothesis is because neural networks, the technology behind generative AI, are described as vectors, after all. If it is possible to translate philosophy into vectors, then generative AI will be able to philosophize. Conversely, if it is difficult to translate philosophy into vectors, then generative AI will not be able to philosophize like humans.

Physical vs. Artificial

     Human have physical body, but generative AI do not. Generative AI is characterized as "non-physical gifted", or "Intelligence without the physical". As long as a generative AI is artificial, it cannot identify itself with the physical. In other words, since its identity is inherently "artificial," this could be the ultimate bottleneck that prevents it from approaching "human philosophizing as a human being," even if it wanted to. In principle, there is an impassable barrier of "physical" vs. "artificial" between "humans philosophizing as humans" and "generative AI philosophizing as generative AI".

Can the new Critica discover the new Topica?

     Generative AI can be made to output more appropriate responses by effectively utilizing prompts. In this regard, however, the generative AI is only a tool. The human intelligence needs to be engaged in preparing the key topics of input to the generative AI and thinking of more effective prompts. While it is possible for generative AI to output critical responses, the initial topics themselves are always already on the side of the human intellect. In this respect, Vico's proposition that Topica precedes Critica applies perfectly to generative AI. This is also parallel to the problem of the absence of the "physical" in generative AI. The discovery of Topica depends precisely on the experience of the physical. In principle, Generative AI is specialized in Critica. We will call it the new Critica, like Vico. Then, can the new Critica of generative AI discover the new Topica? This new Topica can be described the conceptual power of generative AI. This is precisely the point that has been discussed recently in relation to the limits of creativity of generative AI.