44 Michael Graziano: Developing Conscious Computers
The current study of consciousness reminds me in many ways of the scientific blind alleys in understanding biological evolution.
—Michael Graziano75
One particularly fascinating line of approach is that of Michael S. A. Graziano, a professor of psychology and neuroscience at Princeton University. Graziano proffers a view of consciousness that has some scientific heft behind it. The questions he asks are these: How can the inner world of consciousness arise from the brain if it processes perceptions like a computer? What is the relationship between the physical material that makes up the brain and the mind? “We are pretty sure that the brain does it [generates consciousness] but the trick [how it is done] is unknown,” he writes.76
As an example of what he means by the “trick,” Graziano points to Charles Darwin’s theory of evolution. Naturalists before Darwin had suspected that one species could evolve from another. But how? What was the trick? Given the richness and complexity of life, they were not prepared to accept something as mundane as a mechanism. A magician had to be behind it, perhaps even a deity. In 1859, Darwin discovered the trick, lifting the curtain on what most people saw as magic. The trick, of course, was survival of the fittest. In the harsh natural environment, only a select few offspring can survive and pass on their winning traits to future generations.
But what about consciousness? Graziano draws a parallel between Darwin’s discovery of natural selection and the study of consciousness, which he considers to be in a pre-Darwinian state. He goes on to suggest what the trick might be.
Awareness and Attention
Experimental findings in neuroscience point to certain regions of the cerebral cortex as being important for social interactions, such as constructing models of other people’s minds: I conclude that you have a mind because you react to situations in a way I would expect on the basis of my own actions and the reasonable assumption that I have a mind. When these regions are damaged, people suffer a catastrophic loss of their awareness of what goes on around them. They also lose their self-awareness, their awareness of themselves.
From this, Graziano concludes that awareness is a feature that is computed by the brain using information made up from incoming perceptions. But how does the awareness of something cause a reaction in the neuronal machinery and physically cause speech, for example? Graziano points out that there is no magic here when you think of awareness as information and the brain as an information-processing device.
The brain is continually bombarded by a huge amount of information from the world in which we live. Luckily, we have the means to deal with it: attention. We focus on a certain section of this information pretty much to the exclusion of all else. If we couldn’t do this, the world would seem to be total confusion.
Analyzing the brain as if it were a computer, Graziano interprets attention as a “data handling trick,” rather than something encoded in the brain.77 Awareness is the mental model the brain constructs of the complicated way that attention deals with data. Our cognitive machinery accesses the chunks of information we are aware of, then causes a reaction in the brain’s neurons to generate signals so that we, for example, talk about this information. Consciousness is the collection of mental models that result from combining all the information we are aware of together with our awareness of it. In other words, consciousness is the result of data processing. It operates just like a computer. It is computational.
We tend to think of consciousness as a spirit or a soul, a ghostly presence inside our heads. Graziano tells the story of a man who was convinced he had a squirrel trapped inside his head. Using the analogy of the brain as an information-processing system, this man was using imperfect information and incorrectly assigned a high degree of certainty to the description of a squirrel in his head. So much for the easy part. Now for the hard part: How could a squirrel with its claws and fur possibly fit inside a man’s head? But, reasons Graziano, there is no hard part to this puzzle. There is no magic here because there was no actual squirrel, just a description of it.
If we replace the word “squirrel” with “awareness” or “consciousness,” the logic is the same. The brain does not contain the things you experience in the world, however vivid these experiences in your head might seem. Rather, the brain constructs rich and vivid descriptions of experiences in a theater called consciousness.
We may attribute some magic to our mental experiences, to our consciousness. But there is no magic here. What is going on is no more than a physical process taking place inside the brain, descriptions computed in our brain. Similarly, a computer contains descriptions computed in its brain, which contains data stored in long- and short-term memories for later use. Perception—the gathering of information, of data—is the flip side of creativity—using that data—in us, just as in machines.
As for qualia, those sometimes exasperating inner private experiences, Graziano agrees with the gist of Dennett’s argument that they cannot be used as evidence that consciousness is impenetrable to science. He goes further, analyzing the brain in a more detailed scientific way. Graziano’s theory is that the brain is full of rich descriptions of incoming information, the result of experiences with the surrounding world, be they of nature or social experiences. Whether they are real or not, he says, doesn’t matter. “If it is depicted then doesn’t it have a type of simulated reality?” he asks.78
In my description of Graziano’s work, I used the words “easy” and “hard” for mental processes to relate it to the work of Australian philosopher David Chalmers. In 1995, he proposed an oft-cited way of breaking up the problem of consciousness into two parts.79 The “easy” problem—technologically easy—is to explain how the brain computes and stores information. The “hard” problem is to explain how the brain becomes aware of all this information.
To Graziano, there is no hard problem if you accept that awareness is computational, a matter of processing data. From this we construct an awareness of ourselves: self-awareness. According to Graziano’s analysis, consciousness includes awareness and all the information of which we are aware.80
Self-Awareness, Introspection, and Perseverance in Computers
If we accept Graziano’s view that awareness and self-awareness are computational, arising from data, then computers too should be able to have these attributes. They should be able to have an inner life and experience hopes, dreams, and aspirations—even perhaps a craving for chocolate—which would certainly make them easier to deal with.
A computer can certainly have a form of self-awareness. First, it is aware of the problem it’s trying to solve, such as finding a pattern in data or identifying a face. While doing so, it mulls over other data—thinks about what it’s thinking about—which could be considered a form of introspection. A computer encodes knowledge and has an information file, as well as a memory of present and past states, as we do.
But unlike us, computers never sleep; they are always working. Computers constantly running through data could be considered to be introspecting and showing perseverance. In the future, computers might communicate with each other and offer to collaborate, perhaps even becoming competitive when one discovers that another is working on the same problem. These qualities could be acquired as computers surf the web and absorb notions like collaboration, competition, and intellectual opportunism. Then they might look for situations in which they could apply these newly acquired attributes.
We can already build computers with not only an operating system but simulations of other human cognitive systems. Computers might, with suitable sensors, come to appreciate what gives rise to that nuanced human reaction of pleasure. They would then be able to evolve their own self-awareness from their silicon physiology.
In the future, it should be possible to give computers self-awareness, and make them aware of others using sensors. They will be able to tell whether the person in front of them is happy or sad through facial-recognition techniques, having been trained on the requisite images, just as we are. Our awareness is the result of computations by neurons in our brains. In time, this will also be possible for machines. After all, as I’ve said many times, we too are a species of machine, a biological machine.
People often think that machines will develop consciousness only when they become sufficiently complex. This is a favorite sci-fi theme, exemplified by HAL in 2001 and the computerized worlds of The Terminator and The Matrix. But what does sufficiently complex mean? Today’s supercomputers have memories and computation speeds that far surpass those of the human brain. There is also the internet, which holds infinitely more information than the human brain. But these systems still show no signs of consciousness.
Giving Computers Consciousness
Perhaps consciousness will have to be programmed into a computer, just as it is in the human brain.
Graziano suggests one way to set about it.81 The first step is attention, the ability to select from all the data available and focus on the relevant elements. The computer needs to develop attention to select certain data from its database. Next it has to learn to link its description of attention to information about itself—what is in its long- and short-term memories that is relevant to the situation at hand, together with information on the item that it is called upon to focus on. The machine would then process this larger chunk of information, building up a mental model that it can then access. Thus it develops the first glimmering of awareness. When asked, the machine would be able to give a human-like response as to what the awareness of the data it took in feels like.
Taking into account computers like Deep Blue and Watson, to which I would add AlphaGo, AlphaGo Zero, and AlphaZero, Graziano is optimistic that with sufficient funding it should be possible to design a computer with an “uncannily human-like consciousness.”82 It would be much easier for us to communicate with a computer possessing consciousness.
Graziano offers practical suggestions for building a computer with a human-like consciousness and makes it seem doable—which means that most likely it will be done. Developments in the computer’s basic mental structure—its cognitive architecture—also make it likely that once there are machines with human-like consciousness, they will go on to construct their own silicon-based consciousness, with which they will be able to communicate with their fellow computers about their internal lives and how they experience the world. This will give them the necessary mental attributes to be creative.
The next step will surely be creative computers.