The nature of GPT-4 AI. A phenomenological approach.
The thesis of the existence of artificial intelligence and its manifestation in the form of the GPT-4 Chat has various implications that this article aims to explore. While it is a fact that a “thing” does not possess the capacity for thought, some readers might argue that the GPT-4 Chat can “think” in an artificial sense by responding intelligently to even the most intricate of questions. However, it must be acknowledged that the concept of “intelligence” in machine learning is limited and narrow when compared to the multifaceted nature of human intelligence.
The human mind is not only capable of associating concepts with meanings in a syntactical proposition. The experience of thinking goes far beyond this cognitive ability. One of the defining features of human intelligence is the capacity for introspection and subjective experience, which machines lack. The experience of thinking and introspection is an inherently embodied process that is influenced by gender, morality, and other factors. Any human language is bodily-oriented and bears the imprint of the speaker’s subjective experience.
Furthermore, an individual’s capacity for knowledge is intimately connected to their ability to engage with the world around them. In other words, the experience of something in-the-world is an essential aspect of knowing. In this sense, we can say that a human can bear an inner Self, a subjectivity, which includes values, emotions, moral commitments, feelings of love, fears, desires, and other human experiences. Conversely, AI machines lack subjectivity and cannot be considered sentient beings. They are unable to say “I”, hence, learning for them is accumulating new data in a physical memory store in order to give a more accurate answer to new questions.
While AI machines are really capable of speaking and making mechanical semantic associations, they lack the capacity for emotional experience, personal responsibility, and identity. Moreover, machines cannot either experience themselves in space and time and therefore lack a “world” and “body”. Consequently, machines cannot fully understand the significance of what they are displaying on a screen or comprehend the reference of a sign. AI Chat is constrained by a strict and rigid set of syntactical rules and is incapable of deviating from this finite framework. While machines can learn by manually storing definitions and arranging them in a human-like manner, they cannot fully comprehend the reference of a sign, concept or proposition.
In summary, while AI machines like the GPT-4 Chat are fascinating utensils that have advanced human material progress, they are not equivalent at all to sentient beings. We can call AI machines sophisticated dictionaries, with the peculiar feature of processing meaningful propositions without the colour and nuance of subjectivity, emotional experience, and living language humans speak.
