Integrated Information Theory - If correct could humans create artificial consciousness?

Question

First off please keep in mind I am self-learning and am learning about this for fun, I have no end goal. I'm trying to make predictions about what I am learning implies or means, so I can ask better questions and clarify what I should learn/focus on learning. I know I need to spend tons more time in understanding the physics and biology that we currently understand. I'm looking for feedback that can help me either confirm my understanding or correct it and/or point me in directions that will help me better develop it so I can ask better questions in the future.

This assumes IIT is a valid theory.

It seems to me IIT implies that consciousness not only requires integration, but in order to achieve integration the system must be dynamic and able to respond spatially to feedback within the system. As I understand it our neurons grow and die, they can move, they receive inputs from their connections but can also be influenced by molecules that make their way to them indirectly. There is an end result of all the behavior of the neuron and that is the movement of electrons. The molecules and cells surrounding the neurons, provide it the ability to release the action potential. Or the stored up energy that must reach a certain threshold in order for the neuron to fire. This firing is what must be in harmony and certain firing patterns in theory results in conscious experience.

So if humans were to artificially generate consciousness as we experience it (using non-biological methods) this would require us to develop a system that can arrange and rearrange neurons or nodes spatially that can release action potentials of electrons(not sure I'm using correct terminology/physical process) in symphony. The neurons would need to be able to change connection weights not only through internal inputs but also through external inputs, and not only by connected nodes. We would need to understand what causes neurons to grow along particular paths or make particular connections... I assume some of this is known or theorized about. It also seems that certain parts of neuronal intelligence is crystallized or more or less unchanging where as other parts do change and use relationships between the unchanging or slow to change parts in conjunction with more dynamic components that allow for variation in abilities to connect.

IIT seems to imply that conscious experience lies in certain configurations of the electron field (what happens beyond that is unclear) but that seems to be the end result of the actions that any particular group of neurons is taking. I suppose not all, because it also serves the purpose of communication and coordination through out the body. It doesn't seem that synchrony in neuronal behavior coordinates chemical arrangements that would explain experience, but it does coordinate electrical fields as an end result. I'm trying to understand if that is a correct understanding whether that is one of the main outputs of neuronal firing.

Answer 1

IIT implies that just about everything is 'conscious' (to a degree) in that you can measure its integrated information to say how conscious it is, while accepting that what we as humans typically describe as 'consciousness' is a very high value of this measure. That's sort of the whole point of the theory: having a definition of consciousness that starts from a more basic principle ('integrated information') rather than starting at "humans are conscious" without defining what it really means to be conscious.

High levels of integrated information are not easy to achieve with typical computers we use today, but nothing about the theory says it cannot be done. If you have a sufficient level of integrated information, we can call that consciousness based on IIT.

IIT does not say anything about implementation of such a system, even though your question implies that it does (you could view this as either a weakness or a strength depending on your perspective); there are certainly ways to connect artificial systems dynamically without them needing to be neurons. It is true, however, that an individual neuron is far more complex than typical artificial instantiations of neurons, like the nodes of an artificial neural network.

As far as your paragraph beginning "IIT seems to imply that conscious experience lies in certain configurations of the electron field" I don't know what this means or where your got it; IIT doesn't say anything like this. IIT says that consciousness is what you get when you connect information to other information; for example, placing information you are measuring from the world right now in the context of your life's experience involves integrating information from your sensory systems with your memories. Computers can already achieve this to some degree, the part about human consciousness that is remarkable is just how deeply and rapidly (and automatically) these comparisons occur in parallel.

---

Koch, C., & Tononi, G. (2008). Can machines be conscious?. Ieee Spectrum, 45(6).