There are a lot of processes in this world that are computationally irreducible (Wolfram sense). Consciousness is perhaps one of them, but this shouldn't imply that anything that is computationally irreducible is the same as Consciousness.
IIT employs its own idea of an irreducible measure. I suspect that it is ill-defined. But nevertheless, it is senseless to define Consciousness on an ill-defined measure. Defining something based on something else that is ill-defined gets you nowhere.
It is also reckless to begin one's investigation of consciousness based on introspecting one's feel of consciousness. This is because the biological motivation of having consciousness is to deceive the mind into believing the importance of survival.
Living things, with consciousness, have no desire of being terminated. So consciousness must therefore involve the affordances that motivate an agent to keep alive.
Affordances are illusions that drive an agent towards actions that ensures its survivability in its niche. Consciousness is nothing more than a kind of affordance that varies in sophistication depending on the individual.
At Wittgenstein has remarked, even if a Lion could talk, we could never understand it. Analogously, we can never understand how it feels to be an Octopus. Yet, all of these have consciousness. Albeit working very differently from ours.
That said, it's important to study the unique consciousness that is available to humans. This is because our general intelligence is intrinsically entangled with the kind of human consciousness we possess.
We are generally intelligent as a consequence of the affordances that we have through nature and nurture. It is these affordances that shape our perception of our world and thus our consciousness.
The most promising approach to understanding consciousness is thus from an evolutional and developmental perspective. What this implies is that we must understand how the mind creates itself from within.
Ultimately, the reason why machines do not have consciousness it because they are not built in the same way as biological creatures. That is, by incrementally building themselves from within. That is, an inside-out development plan.
An inside-out development plan is that everything that is generated is grounding in this reality. That is, as an organism builds multiple levels of complexity, each level must learn about its environment. It learns judgment across every scale.
This is Cantwell-Smith's argument why AI can never achieve general intelligence. This is because AI has no judgment about this world. It has no 'skin in the game'. It does not know what is relevant having never been born into this world.
So we've come full circle to explain why formulations like IIT that are defined outside-in can never be an explanation of consciousness. Human descriptive models are not models that have semantic grounding. It is only generative models that are semantically grounded.
The thing though about generative models is that they can be computationally irreducible.
So this is my explanation of consciousness in a tweetstorm.
Let me know if you find any holes ( or why it is whole).
@threadreaderapp unroll
