As explained last post, in regards to consciousness I’ll assume a materialist biology theory of mind. And as such the science question is how a real human brain is conscious. This approach avoids any abstract philosophical arguments such as the mind-body problem. Furthermore I’ll immediately concede I’m in the tank for Daniel Dennett, pictured above. So to a certain extent this post will be a commentary on Dennett’s ideas, with a little Jeff Hawkins on the side.
So starting from materialist theory of mind, let’s develop things in the order below:
- Jeff Hawkins – mind as memory prediction framework
- John Searle’s Chinese Room problem
- Determinism and Free Will
1. Hawkins – Mind as a Memory Predictive Framework
Hawkins book On Intelligence was published in 2004. His co-author was Sandra Blakeslee, who also co-authored books with the excellent V. S. Ramachandran. Now Hawkins is actually interested in building intelligent machines, but to get there he first walks through how the brain works. In Hawkins view, the brain and nervous system is a hierarchical memory prediction framework. Let’s unpack that phrase by starting first with prediction. The nervous system continually tries to predict what it expects to sense next. Then it predicts how any proposed action changes what will happen. So maybe the brain predicts what you’ll feel next as you roll your keys around in your hand. Or maybe it’s whether a ball heading toward your eye will hit you in the face. The hierarchy comes in by how predictions flow up and down. Lower level predictions (blinking will stop sudden brightness) are fed back into higher level predictions (sun is in shining in your eye), which then flow back down (predict if you move your head your eyes won’t hurt anymore). This is based in particular how the neocortex is constructed. Finally memory comes into play because memory lays down the basis for future predictions, and in particular the self correction needed to adjust to the world. So if you practice catching a ball, what happens is the entire memory prediction framework predicts where the ball will be next and what happens as you move your hand toward it. With repeated practice the error between prediction and what happens gets decreased due to automatic adjustments. Eventually you learn to catch a ball. While the actual calculation mechanism and substrate are utterly different, the analogy of how a brain learns and how a predictive statistical error algorithm works is valid. Both require massive data sets to slowly learn.
Now we’ll add Dennett into this and define consciousness as the ability to predict the results of your possible actions, along the Hawkins model. And the highest form of consciousness is being able to predict other people’s actions. That is, having a theory of mind. We are evolved to do this. And at least to some extent you can rank levels of consciousness (say among animals) by how well they can predict how their actions will impact the world around them. With social animals of course having an edge since they can predict actions of others (theory of mind) since that’s required to interact socially.
2. Searle – Chinese Room problem
John Searle came up with the famous Chinese Room thought experiment in 1980. Suppose an AI can pass the Turing test in Chinese. It can take any written question in Chinese and give a written Chinese reply like a human. Then replace the computer with a human who only knows English, and this person follows the machine algorithm only with pencil and paper but eventually returns all the same Chinese answers as the computer would. Clearly the human doesn’t “understand” the Chinese conversation.
Now I love the quote from wikipedia on this topic where they start by saying this is the most discussed problem in cognitive science for the past 25 years. So we can’t avoid it here. But then they go on to quote: ”The overwhelming majority,” notes BBS editor Stevan Harnad, “still think that the Chinese Room Argument is dead wrong.” One solution is to realize that consciousness is measured at the aggregate level, so the human inside the computer is by analogy like an individual brain cell. But more importantly from a Dennett/Hawkins point of view, the reason this is wrong is that mind is about predicting how your actions impact the world. So to really be conscious the computer that passes the turing test must also be able to interact in the world and predict actions of other thinking beings. The written turing test is not complete. Having a robot walk around and interact in the real world is the real test of an AI.
3. Qualia – real but not impacting main argument
We have to cover qualia since it’s often discussed as a key issue for understanding consciousness. Qualia are the subjective experience of the smell of a rose or the color blue. Dennett’s take is qualia don’t exist. Here’s one where Dennett misses the mark and his critics like Ramachandran rightly hammer him. But even if you accept that qualia exist, the stronger arguments from people like David Chalmers arguing qualia are required for consciousness don’t follow at all. That’s clear once you take point #1 above as valid, where consciousness is defined by the ability to predict the impact of your actions.
4. Determinism and Free Will – compatibalist view
In regard to free will, we should start by pointing out that most philosophers are compatibilists. That means they believe determinism and free will are compatible, as David Hume argued. This is Dennett’s view as well of course.
My spin on why some people don’t agree with compatibilism is they get caught in a simple fallacy. The fallacy is to argue that just because a choice is predictable it is not a choice. This is a fallacy. If you can predict I will duck my head when you throw a brick at it, that doesn’t prove I didn’t make a decision to move my head. If you go back to the memory-prediction framework in point #1 this becomes more clear. A brain works by predicting what will happen in the real world, and then predicting how actions will alter outcomes. Whether people’s choices are predictable is neither here nor there. The choice still happens in the brain even if it’s a predictable choice. As such people are accountable for their choices, as long as their brain is not damaged so they can manage this process in a normal human fashion.
So why discuss consciousness and free will at all if it’s settled?
Doing a quick google search on “free will” and “mind”, I get 15 million results. After wikipedia, I get results like:
- Scientific American – “Scientists say free will doesn’t exist….”
- Scientific American – ”Neuroscience Challenges Old Ideas about Free Will”
- “The topic of free will has challenged thinkers and inspired debate across multiple disciplines for centuries.”
- “Do you believe in Free Will? Maybe you should even if you don’t.”
Just eyeballing it I would guess at best half support the common sense view that free will exists. But the bigger issue is that even the articles supporting free will treat it as something that’s not settled. Reading the press on free will or consciousness make me feel like I’m trapped in a giant dorm room with a bunch of stoned college students. Dude – if I have one more toke, is it my mind making me do that, or is it my brain? Can I even make a choice since you know I’m gonna do it? It’s mindblowing to think about this stuff dude.
One reason this “debate” exists is the media gets good headlines by saying free will doesn’t exist. But the deeper reason it exists is because it’s a proxy for a more fundamental debate: whether a materialist view of mind is true in the first place. This is especially true since spirituality is tied to the belief in more than just the material world. But if you do happen accept that materialist view of mind, there’s really not too much to talk about except working out some details. If science, then free will.