Consciousness is a WHAT kind of loop?

The new theory that explains the what, why, and how of consciousness as a computational meta-management process

Three questions of consciousness (Courtesy of DALL-E)

What is consciousness? Ever since René Descartes first proposed the Dualistic theory of Mind and Body, people have been trying to answer that question, and yet we’re still arguing just as much as Descartes and the Princess of Behamia. The more that have I read about that question the more I have noticed a trend: that discussions of what consciousness is usually ignore the questions of why it’s there in the first place or of how it might function, and conversely that discussions of why it’s there and how it works usually ignore trying to define what it is. This leads to some strange and other-worldly ideas, and to some gross understatements of the problem at hand.

All three questions are difficult to answer. Answering all three questions together must surely be harder right? Certainly, that’s the assumption that has led to most of the divide-and-conquer style of discussions so far — make some vague assumption about two of those questions and try to answer the third.

I believe we already understand consciousness. The problem is to see that understanding in amongst all the conflicting ideas.

In my latest report I intertwine discussions to all three questions and I think it yields some answers that haven’t been stated quite as clearly before.

You can read the full report here (it’s just as free as this post):

Consciousness is a Semiotic Meta-management Feedback Loop

Those familiar with the area will notice I don’t raise many points that are entirely novel on their own. What is novel is how I’ve pulled those points together. I actually believe we already understand consciousness. The problem is to see that understanding in amongst all the conflicting ideas.

I’ll do a series of posts breaking down the ideas in time. For now, here’s a summary of the key points…

Why

Something has to monitor and train the ability to just do a loop. Enter meta-management.

All that subjective first-person experience of the world and ourselves, all that feeling of a “little person” in the head, all of that “it’s so mysterious it surely can’t exist as a physical process”. That stuff. It didn’t evolve for its own reason. It wasn’t even the key thing to evolve.

Evolution needed to discover a way of making more and more advanced inferences in the face of more and more complex environments and interactions (particularly with other individuals). At some point this led to the evolution of processing loops — yes, just like the sort that we have in programming languages.

Anyone who’s done work with artificial neural networks in reinforcement learning settings knows that they lead to pretty chaotic behavior until they’ve undergone a lot of training. The processing loops I’m talking about are governed by those same chaotic (biological) neural networks. Something has to monitor and train the ability to just do a loop.

Bee watcher watchers watching the bee watchers (Courtesy of DALL-E)

Enter meta-management.

It’s a second (higher-order) process that watches the first-order process, and modifies it if needed.

In the full report I explain in detail how such a higher-order process can work, and how to avoid the infinite regress of another bee-watcher-watcher watching the first bee-watcher.

How

This leads to higher-order thoughts (HOTs) that are generated about the goings on within the process, and which are also used by that same process to examine and manipulate itself.

Bee watching itself (Courtesy of DALL-E)

In a strange turn of events that would make Douglas Hofstadter and his I am a Strange Loop proud, it turns out that a process can meta-manage itself — probably, under the right conditions.

I propose that that is possible under two conditions:

• the right kind of feedback loop, and
• the right kind of evolutionarily hard-wired constraints on learning.

The feedback loop is what makes our subjective experience of ourselves seem like another perceptual sense — because it is one.

But it can’t just feed an exact copy of our brain state in as a sense. There’s too much going on (and another infinite regress to contend with). It must be a higher-order summary.

This leads to higher-order thoughts (HOTs) that are generated about the goings on within the process, and which are also used by that same process to examine and manipulate itself.

With a layering of evolutionary hard-wired unconditioned responses, plus conditioned responses, plus learned habitual behavior, plus higher order executive function, it’s possible for the process to simply discover the benefits of using that “meta-management feedback loop”.

Meta-management is thus an emergent process.

What

Consciousness, including everything about how mysterious it feels, follows from computations over that data.

So, what is consciousness?

Consciousness is a by-product. It’s a result of the meta-management feedback loop generating a summary of key goings-on within the brain and providing that back into those same processing systems. It’s a result of our elaborate abilities to form interpretational models of our world (both the external and the internal world), and to use those models in interpretation of our perceptions.

Consciousness, including everything about how mysterious it feels, follows from computations over that data.

There’s all sort of fascinating things to come out of this. Like why only certain brain processes are consciously accessibly. Or a resolution to debates around the causal nature of consciousness. And whether it’s possible to create artificial consciousnesses within a silicon von Neumann computer.

For now those issues will have to wait for a later post.

Here’s the link to the full report again:

Consciousness is a Semiotic Meta-management Feedback Loop

What do you think? I’d absolutely love to hear your thoughts and questions. So grab a coffee, have a read of the full report, and please do leave a response.