In recent years, the development of Artificial Intelligence (AI) has sparked discussions about the potential for these systems to possess consciousness. However, it is possible that we may be underestimating the neurobiological mechanisms underlying human consciousness. Modern AI systems exhibit […]
In recent years, the development of Artificial Intelligence (AI) has sparked discussions about the potential for these systems to possess consciousness. However, it is possible that we may be underestimating the neurobiological mechanisms underlying human consciousness.
Modern AI systems exhibit remarkable behaviors. For instance, when we interact with systems like ChatGPT, the responses can sometimes closely resemble those of a human being. As we communicate with ChatGPT, we consciously experience the text it generates. In fact, you are consciously experiencing this text right now!
The question arises as to whether the language model itself is conscious of our text prompts or if it simply functions as a pattern-recognition algorithm. Based on the text it generates, it is easy to think that the system might be conscious. However, in a study conducted by Jaan Aru, Matthew Larkum, and Mac Shine, they approach this question from a neurological perspective.
All three authors are neuroscientists, and they argue that despite the appearance of consciousness in systems like ChatGPT, it is highly unlikely.
Firstly, the input to language models lacks the embodied, contextual information characteristic of our sensory interaction with the world around us. Secondly, the architectures of current AI algorithms lack key features of the thalamocortical system associated with conscious awareness in mammals. Lastly, the evolutionary and developmental pathways that led to the emergence of living organisms likely do not have parallels in today’s artificial systems.
The existence of living organisms depends on their actions, and their survival is intricately linked to multi-neuronal, intercellular, and organism-level processes that culminate in agency and consciousness.
Therefore, while it may be tempting to assume that ChatGPT and similar systems could be conscious, it would be a serious underestimation of the complexity of neural mechanisms that generate consciousness in our brains.
Researchers disagree on how consciousness arises in the human brain. What we do know, and what this new study suggests, is that the mechanisms are likely far more intricate than those underlying current language models. For example, as stated in the study, real neural cells are not the same as the neural cells in artificial neural networks. Biological neural cells are physical entities capable of growth and shape changes, whereas neural cells in large language models are merely insignificant pieces of code. There is still much to learn about consciousness, and consequently, there is a long road ahead towards conscious machines.