Robotics, Artificial Intelligence and St. Thomas Aquinas

Could futuristic technology ever have a soul?

By Gabe Ferrer




S
ince seeing "Star Wars" as a five-year-old boy when it first came out in 1977, I've been fascinated by futuristic technology. From that time, I was pretty sure I either wanted to create intelligent robots or develop a hyperdrive for faster-than-light travel. 

As things turned out, I've developed a passion for mobile robotics. As a professor, I have the opportunity to teach one to two courses per year involving robotics, in addition to my research on robotics topics. A natural question that arises when thinking about robots is whether a robot could ever be a "person", or have a "soul". Could a robot ever be equal in status to a human being? Would it need equal rights? 

The Star Wars movies themselves exemplify the tension between these alternatives. R2D2 and C3PO are definitely portrayed as distinct personalities, with human qualities such as courage and fear. On the other hand, they clearly do not have human status; they are bought and sold, and no compunctions are held with regard to erasing their memories. Given present technology, I do not think anyone would be concerned about any contemporary robot exhibiting personhood. 

On the other hand, something about watching a mobile robot in action affects us in a psychological way. My students are always eager to name and personalize their robots. I even once heard a story about an insect-like mine-sweeping robot being demonstrated for an army officer. Every time it touched a mine, it lost a leg in the subsequent explosion. After having only one leg left, allowing it to crawl rather pathetically, the officer ordered the demonstration stopped on the grounds that it was "inhumane."

To think more deeply about this issue, we have to examine just what a robot is. Simply put, a robot is nothing more than a computer with sensors and motors. To program a robot involves specifying what the motors should do given certain types of information from the sensors. Imagine a robot with two motorized wheels and a castor wheel. It is like a backwards tricycle. If this robot also has a sonar sensor mounted in front, it is easy to write a program telling it to turn left if the sensor detects an object within one meter, and that otherwise it should drive forward. While programs of much greater sophistication are certainly possible, the essence of a robot control program can be seen even in this simple example. 

The robot is programmed to make decisions based on sensor values that have been turned into numbers. That's all there is to it. Consider a robot with a webcam, that makes decisions based on video input. To a robot, a video stream is nothing more than a sequence of frames, as if you were watching a movie. And each frame is nothing more than a giant grid of numbers, with each number representing a color. 

But it is admittedly tricky to come up with a simple rule for saying whether or not a given image corresponds to an area the robot can drive through. One common way to deal with this is to create what is called a "neural network." In the brain, there are groups of neurons that will be active if, say, you see a door. A different group may be active if you see a flower. A neural network simulates this. By showing examples of different types of objects, different parts of the neural network learn to become active for each of the object types. The use of the term "learn" is loaded, of course. 

What really happens is that, just as the images are just a bunch of numbers relative to a computer, so likewise is a neural network. The learning process, then, involves doing a lot of math to modify part of the neural network to make its numbers correspond to the numbers comprising the input image. This brings us to St. Thomas Aquinas. For Aquinas, there are three kinds of souls that exist in living beings: vegetative, animal, and rational souls. A vegetative soul is capable of processing nutrients, growing, and reproducing. An animal soul adds to this sensation, locomotion, and appetite. A rational soul has a will and a capacity for abstraction.

If Aquinas had been aware of computers, robots, and neural networks, he would have denied that a machine could ever have a rational soul as a human does. At most, a machine can simulate the sensation and locomotion aspects of an animal soul. It would not even, properly speaking, be able to simulate appetite. A robot could be programmed to seek out a wall socket if its batteries are low, but it is an altogether different type of reality than an animal seeking food. 

And if a robot cannot even be equivalent to an animal, how much less would it be equivalent to a rational human being. Leaving aside the issue of free will, Aquinas observes that humans have a capacity for forming abstractions. What a neural network does, or even what an animal does, is very different. A neural network might encode an example of, say, a ball. This encoding is effectively the result of averaging together all the numbers corresponding to all of the balls the robot has ever seen. It is all purely material. 

But humans can form immaterial abstractions. We can conceive of the idea of a "ball" beyond a particular ball. For that matter, we can use introspection to imagine what someone else is experiencing. Aquinas argues that this capacity of the rational soul is one way that we know that the rational soul cannot be entirely material. But any neural network or computational data structure will be purely material. To encode my concept of what my reader is thinking as he reads this essay as a bunch of numbers is a conceptual impossibility. This little essay is obviously not the last word on the subject.

Much ink has been (and continues to be) spilled about these issues by some very busy philosophers. But what really amazes me is how the resources of the Catholic tradition (in this case, the philosophy of Aquinas) have so much to offer in thinking about even the most cutting-edge of scientific issues.

Posted in: Science, Technology