How to Create a Mind: Book Review
Could it be possible to create a digital mind? Ray Kurzweil thinks so. In his book How to Create a Mind: The Secret of Human Thought Revealed, he sets out to demonstrate how this could work.
I have to start by saying that Kurzweil is not a philosopher nor a neuroscientist, he is a futurist and computer scientist who has worked with optical character and speech recognition software, among other endeavours. He says, for example, that his research and software were later used in a software called Dragon Naturally Speaking and more recently in Siri.
That alone doesn’t qualify him to write a book about the mind. However, by reading the book you can tell that he has studied the subject of the brain extensively.
In the book, he lays down a framework he calls PRTM (Pattern Recognition Theory of the Mind). It is a similar theory to the one exposed by Palm Pilot inventor Jeff Hawkins in his book On Intelligence, but with some differences and additions.
I will not give a full explanation of the PRTM here, first because I don’t fully understand it, and second, it would be too long and useless for our purpose.
What I will say is that the way our brains work is relatively simple and straightforward: according to Kurzweil we have about 30 billion neurons, but these neurons are grouped in about 300 million of what he calls neural pattern recognizers, so each neural pattern recognizer has about 60,000 neurons.
Each of these pattern recognizers store part of our world, and sequences of patterns recognizers are what we call memory. There is a lot of redundancy of information in these pattern recognizers. Also, pattern recognizers that are not used enough get recycled to store other things (that’s why you can sometimes remember parts of the events you experience).
Saying that these pattern recognizers are responsible for all of our perception, memories, experiences, actions, desires, dreams, etc. can be seen as simplistic, but I don’t think so. That would be like saying, for example, that matter cannot be made up of molecules, or computers from logical gates, because that’s too simplistic. But the truth is that the PRTM is just a theory, we really don’t know for sure yet.
Although the PRTM theory is the core of the book, it wasn’t what I enjoyed most for a number of reasons. First, I had already read On Intelligence, which lays a similar theory about the brain but much more detailed. Second, although studying the brain is important per se, I don’t think that it can help us in our AI efforts. Or to put it in another way, we don’t need to understand how the brain works to create a digital simulation of the mind.
I also think the title is misleading and too ambitious. I think that we could eventually simulate a mind to the point that we won’t be able to differentiate it from a real mind, but it will still be a simulation, it would be a program that “acts” like a human mind. We are lazy, we experience emotions, we have a huge statistical bias and we are not rational. We could simulate all this, but it would still be a simulation.
After exposing the PRTM theory, Kurzweil explains the research that is underway, like the Blue Brain project that wants to simulate the brain at the molecular level. Also, he superficially explains Artificial Neural Networks and Hierarchical Hidden Markov Models, which are, basically, computer algorithms. With that, he sets out to describe a strategy to build a brain simulation which, he says, should have a purpose expressed like a series of goals. Here I would argue, and I think he does in another part of the book, that we have already built multiple minds (think of Watson, Siri, self-driving cars, Waze, Shazam, Google, etc.).
The chapter I enjoyed the most is called Thought Experiments on the Mind that has a lengthy discussion about consciousness, free will, and identity. This chapter seems detached from the rest of the book but allow us to understand better why this is a difficult subject and why Kurzweil is willing to say that a digital mind is conscious.
After a lengthy explanation of various positions about consciousness, he offers his own: that consciousness is a property of any complex physical system. So, he says, an ant has some level of consciousness, but less than a dog. In that sense Watson and Siri are conscious. But he accepts that each person has his or her own test for consciousness.
I agree that consciousness is a result of something physical. But my view is that consciousness is something inherently human, the capacity of self-awareness and thinking about thinking. We develop that capacity in the first years of our lifes. The problem is that eventually we will be able to create a computer software (maybe inside a human-like body) that tricks us about its consciousness, so we won’t be able to recognize true consciousness.
To conclude, I think the book is well researched and packed with interesting information, but I would suggest changing its title to something like How to Simulate a Mind: a Theory of Human Thought, which is not that catchy but less misleading.
