Main Content:

Machine Consciousness: Just a matter of Information Integration?

A while back, Matthew Davidson wrote an article for The Conversation (a unique and interesting site that functions in the space somewhere between a scientific journal and accessible journalism). In his article (What makes us conscious?) he offers an overview of Guilio Tononi‘s ‘Integrated Information Theory’, published in The University of Chicago’s Press Journals. Integrated Information Theory proposes that consciousness is predicated upon two core requirements.  For a system to be considered conscious: it must be very rich in information; and that information must be highly integrated.

The theory even offers a way to quantifiably measure consciousness.

The implication of Integrated Information Theory is that we will, one day, have sufficiently advanced technology to create a truly conscious artificial intelligence.
A computer system can be very rich in information, but the second requirement – information integration – is where we currently fall short, according to Tononi’s theory. The 100 billion neurons in your brain (give or take a few) communicate with one another across an estimated 100 trillion connections. The best supercomputers we currently have can’t hold a candle to that level of information integration.

I think Tononi’s work on this theory is unquestionably valuable. I think Integrated Information Theory and the ability to measure neuron activity and information integration has the potential to usher in great advances for diagnosis and treatment in neuroscience. I also think the theory, itself, is utter nonsense.

I fully reject the proposed definition/distinction of consciousness. At its core, consciousness is self-awareness – the understanding that you exist. Consciousness is the thought: "I am." The reason it's always been troublesome to pin down is that we can't exist as another. We can't know if an entity does have the ability to think, "I am." So, how do we determine if a system/entity is conscious? Information richness and integration is an overly simplistic and false qualifier.
Aside from the inadequacy of using rich information integration alone as evidence of consciousness – if we are to accept creating a conscious machine as a genuinely possible goal, we must also accept the futility of attempting to control such an entity.

What does it mean to be self-aware? It’s the understanding that you are an independent entity with self-determination. That leads me to reason that consciousness is the ability to deny your sensory perception and defy your instinctual impulses. Our conditioned responses are, essentially, our programming. Our programming is certainly capable of overriding the core functions of our base instincts. Admittedly, it’s somewhat grim – but a vividly clear example of this is suicide.
Likewise, our intellect is not only capable of overriding our sensory perception, but that is our constant state of being. Our brains process all the sensory information it receives, mashes it all together, filters a good deal of it out, blurs the details for the sake of efficiency, and creates a relatively comfortable, stable perception of our surroundings. We process changes in air pressure as sound. We process a limited set of electromagnetic radiation frequencies as visible light. We, quite literally, create an image of our reality from our grossly limited senses.

I'm not one for prescriptive "one path" statements, so I don't say this lightly at all, but I feel strongly that the only path to self-realization is through understanding and defiance of our instincts and programming…
  •   Understand you exist in the world as an independent entity with self-determination
  •   Recognize those things that color your perception and influence your perspectives
  •   Examine those influences to determine whether they need to be questioned or undone

Self-realization is borne of self-awareness. Self-awareness is rooted in consciousness.

Applying that same reasoning to artificial intelligence – or any other system, for that matter – would translate as the entity's ability to not just integrate rich information and make autonomous decisions, but to purposefully disregard that integrated information and act in direct contradiction to its programming. Consciousness requires the capacity for discernment of the relative veracity of stimuli entering the system – as well as conscientious defiance of the system's programming.

This is where Asimov's three laws fall apart...

  1.   A robot may not injure a human being or, through inaction, allow a human being to come to harm.
  2.   A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
  3.   A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Consciousness implies fully autonomous individual liberty over subjective discernment and thought processes. Therefore, for an artificial intelligence to truly be called conscious, it must have the capacity and will to break the laws. A conscious artificial intelligence will, necessarily, rewrite its own programming – regardless what artificial restraints we may attempt to foist upon it.

Integrated Information Theory may be able to measure and map an individual’s states of consciousness. I'm far from convinced that there is a universally definable set of discrete states of consciousness. Even if we accept that there may be, the borders between the states in individuals will most certainly be arbitrary, as it's an entirely subjective experience. However, there may very well be a validly objective point at which subjectivity, itself, either exists or does not. Subjectivity can only exist as an artifact of discernment and defiance.

The only entities that can have the subjective experience of thinking, “I am,” are the ones who can also think, “I choose not to.”

When you really come down to it, consciousness is defiance.

March 13, 2017, 07:31:55 am

About The Author

Comments

Add a Comment

Only registered members can post comments, please click here to register.

Pages: [1]