Introduction
In my lectures, I always emphasize that it's not just computing power that determines the degree of intelligence. Intelligence is a complex phenomenon that involves not only data processing but also dynamic and contextual interactions with the environment. I recently came across an intriguing article that reinforces this view: "Artificial Consciousness: A Free Energy Principle Perspective," by Wanja Wiese, published in Philosophical Studies in 2024.
The Origin of the Article
Wiese's article explores the possibility of consciousness emerging in artificial systems, analyzing this question through the Free Energy Principle (FEP), proposed by Karl Friston. FEP is a theory that seeks to explain how self-organizing systems, such as living organisms, maintain their internal order by interacting with the environment to minimize uncertainty.
Simulation is not Replication
One of the central questions raised by Wiese is: is the simple computational simulation of neural processes sufficient to generate consciousness in an AI? Or would something more be needed to truly replicate conscious experience?
The Role of the Free Energy Principle
According to the PEL, self-organizing systems minimize "free energy" to maintain their structure and survival. This means they are constantly adjusting their predictions and actions based on interactions with the environment. This dynamic creates a specific causal flow that is intrinsic to living systems.
Causal Flow in Living Systems vs. Computers
Wiese argues that this causal flow, present in living organisms, is not replicated in traditional computers with von Neumann architecture. In organisms, there is a direct and continuous interaction between internal states (such as beliefs and expectations) and external states (such as sensory stimuli). In computers, however, this interaction is mediated differently, which may be crucial in distinguishing between simulating and truly replicating consciousness.
Implications for Artificial Intelligence
If we accept that consciousness requires more than the simulation of computational processes, we need to reconsider how we develop our AIs. It may be necessary to go beyond increasing computational power and create new architectures that can replicate the causal flow and dynamic interactions found in living beings.
What Does This Mean in Practice?
- New Hardware Architectures: Develop systems that not only process information, but also have causal interactions similar to those of living organisms.
- Integration with the Environment: Create AIs that are not isolated, but that interact continuously and adaptively with the environment.
- Rethinking the Nature of Consciousness: Understanding that consciousness may not be a byproduct of information processing, but rather how that processing is integrated into a larger system.
Conclusion
The discussion about artificial consciousness is as much philosophical as it is technological. It's not just about increasing the processing capacity of machines, but also about understanding and replicating the complex interactions that give rise to consciousness in living beings. This leads us to ask: are we on the right track to creating truly conscious machines, or do we need a completely new approach?
