A provocative look at the theological implications of artificial intelligence from the founder of MIT's God and Computers Project.

Get ready to meet two remarkable characters, Cog and Kismet. They both enjoy working with others, they're very attentive, have excellent learning skills, and, according to their colleagues, they're very charming. And they're both robots.

From Hollywood to the halls of NASA, robots loom large in the popular imagination. But what feelings do these lifelike machines really provoke in us? In God in the Machine, Dr. Anne Foerst draws on her expertise as both a theologian and computer scientist to address the profound questions that robots such as Cog and Kismet raise for us all: How do we define "human" versus "person"? What does it mean to have a soul? And what do robots teach us about our relationship with God?

"I have, alas! Philosophy,
Medicine, Jurisprudence too,
And to my cost Theology,
With ardent labor, studied through.
And here I stand, with all my lore,
Poor fool, no wiser than before.
Magister, doctor styled, indeed,
Already these 10 years I lead,
Up, down, across, and to and fro,
My pupils by the nose,-and learn,
That we in truth can nothing know!"

Johann Wolfgang von Goethe,
Faust, Part I

This is the story of a handshake

The handshake occurred in the fall of 1995 between Harvey Cox, a professor at the Harvard Divinity School, and the humanoid robot Cog, developed at the Artificial Intelligence Lab at the Massachusetts Institute of Technology.

Cog is a huge robot, approximately seven feet tall, and somewhat intimidating. It has a massive steel frame. Our culture is infused with science-fiction stories about robots that turn against humans and destroy them. We are also accustomed to factory robots that blindly do their assigned tasks and can accidentally smash a finger or a head. Therefore, Cog induces a mixture of fascination and fear in many people who see it for the first time.But Harvey Cox took the first step. He had devoted his entire life to bringing Christian theology into a dialogue among people with different worldviews, and he didn't want to stop at a robot. So, when he first realized that Cog's eyes were following him around the room, he made eye contact. Then Professor Cox tentatively extended his hand, and Cog, after some trial and error, grasped it. There was a collective gasp from the Harvard theologians and MIT scientists present.

Cog is the result of one of the first attempts to build a humanoid robot, and in 1996 it had just become very famous. As soon as the media discovered the project, Cog was constantly surrounded by cameras and covered by all the major television networks and newspapers. Building a humanoid robot is a fascinating endeavor, but before the development of Cog, such a project had belonged firmly in the realm of science fiction.At MIT, Rodney Brooks, then associate director of MIT's famous Artificial Intelligence Laboratory (MIT AI Lab), had the vision of building humanoids. For quite some time, Rod was the enfant terrible in the robotics community; he had played around with techniques and strategies that were completely new to the field, and he was the first to build successful autonomous robots that could navigate real-world environments. This might sound like a very simple task, but before the mid-1980s, nothing like it had ever been accomplished.

The traditional metaphor in AI had been that intelligence is a program that is implemented in the wetware of the brain but might just as well be implemented in the hardware of the computer. Traditional AI programs have internal world models that process data in order to plan "intelligent" reactions to given stimuli. While these traditional machines are great at playing chess and proving mathematical theorems, they fail as soon as they have to navigate constantly changing environments. They are incapable of reacting effectively to the real world. The assumption in the traditional system is that intelligence is the capability for abstract thought. Humans are so good at it that they can live in the real world. If AI systems become increasingly smart, they might eventually be able to perform as well as we do.

Rod broke with this assumption and declared our facility for abstract thinking a mere by-product of our ingenious capability to intentionally shape our respective environments in order to survive. Rod believed that if AI researchers attempt to build intelligent machines, they ought to build embodied entities that interact with the real world. One might not need abstract thought in the beginning-after all, most animals survive pretty well without it, and human newborns don't have this capability either. Hence, Rod proposed starting with insect robots and then attempting to build a robot analogous to a newborn baby that learns through social interaction and physical embodiment. The jump from insects to a humanoid is huge, but the group that worked with Rod was so excited about the project that they decided to go ahead nonetheless.

At research universities such as MIT, most professors have groups that consist of undergraduate and graduate students, Ph.D. candidates, postdocs, researchers, and visiting professors. That is, people on many different levels work together and inspire each other's projects. I started to interact with Rod and his group in the fall of 1993 and joined the AI Lab in the fall of 1995. While I never got to know all the people in the original project that started in 1992, I became acquainted with all those who worked in the group between 1994 and 2000. At this time, we called our group "The Zoo," as there were so many exotic people in it. As a theologian, I was probably the most bizarre member, but we had other strange animals as well.

I was working in Germany on my doctoral thesis. I had studied computer science as well as theology and focused in my thesis on the possibilities for a nonjudgmental dialogue between Christian anthropology and AI. As I based my work on the theology of the great twentieth-century theologian Paul Tillich, I had intended to work in the Tillich archives at the Harvard Divinity School (HDS) during my first visit to Cambridge. But I also wanted to sneak my way into the hallowed halls of MIT, the cradle of AI. Most bigwigs in AI had come from this place or had trained here, and I couldn't wait to walk the same hallways with these figures that I so secretly admired.

When I met Rod, he invited me to come back in the fall of 1993 to join his informal seminar on "Embodied AI," in which he would talk about his new assumptions on how to build smart machines. I think I tickled his sense of humor when I first walked into his lab. He had probably never been sought out by a theologian. But he respected my wish to analyze the hidden assumptions, hopes, and beliefs in his group, and thought it might help them to develop a clearer worldview. He invited me back again and then invited me to join his lab as a postdoc. After finishing my thesis in 1995, I became an official member of the lab.

Meanwhile, at Harvard, my research in the Tillich archives had not gone unnoticed. I had unearthed a total of six lectures that Tillich had given at MIT, only two of which had been published to that date. Harvey Cox became fascinated. He had studied with Tillich and had written his doctoral thesis on Tillich and technology, but he had never been involved with MIT, nor had he ever concentrated on a particular area of technology. He was especially intrigued by the possibilities of bringing Tillich into the discussion of the development of Cog. As plans for my postdoc time at MIT became more concrete, he managed to get me a position at Harvard as well. This way, I would not spend my entire time at MIT, an alien place for me, but would find refuge at HDS, which was my theological home.

It worked beautifully. I spent my days happily at MIT and often trotted over to HDS. The inherent differences between both places fascinated me and became instantly visible just by looking at the architecture of the respective schools. The AI Lab was then a harsh, concrete, cubicled building at Technology Square, and HDS a Gothic, Oxford-style building with turrets. One had long corridors, chaotic offices, and modern furniture, and the other had oil paintings and woodwork from another era. At MIT, you couldn't distinguish between professors' and students' offices; at HDS only professors resided. Robots were always running through the hallways of MIT, and Cog was constantly being haunted by the media.

But there were also commonalities. Both schools were located slightly off campus and, in both groups, I laughed more than in any of my previous German academic settings. The other commonality was that at both schools I was initially perceived as an outsider. People at HDS found my fascination with high-tech somewhat strange. Most of them were slightly antitech and thought my quest to bring theology and AI together somewhat superfluous. At MIT, on the other hand, people were suspicious of the theologian in their midst. It probably took a year and many, many conversations until heads stopped turning toward me whenever the term evolution was mentioned. Most people there associated any form of Christianity with creationism and had a hard time understanding why I, as a Christian, actively sought out scientific explanations for why we are the way we are. But at both places, initial skepticism slowly evolved into active interest in my research.

As I liked both places so much, I attempted to bring the sides together. Socially, I introduced as many people as possible, and it is wonderful to see that after all these years some friendships still remain. At the same time, I was attempting to bring both groups together academically. I read and gave presentations to prepare them for each other. Finally, the group from HDS came over to MIT to meet the robot team and the robots themselves. It was on this visit that the monumental handshake occurred. There were quite a few important handshakes that day between various professionals of divinity and of AI, most of whom had never met a person from the other discipline.

The handshake between Harvey and Cog was deeply profound, as it built a bridge between these seemingly different areas. When Harvey and Cog looked at each other, it became clear that there was a dialogue waiting for us. As our technical creatures become more like us, they raise fundamental theological questions. As theologians, we have the responsibility not to shrink away from such a challenge but to seek out these opportunities and overcome our fear as Harvey had done that day.

This handshake has been fundamental for my life as a researcher, and I have worked on stabilizing the bridge that was forged at that moment ever since.

This became especially challenging when, after a few years of work on Cog, Cynthia Breazeal, one of the first engineers of Cog, started her own robot project, Kismet. Kismet is the most charming robot I know. Today it is proudly presented in the MIT Museum. However, during my last years at MIT, it was the center of our attention and eventually became the center of my research.

For me, the challenge is to analyze the attempt of the people at MIT to build creatures with human capabilities by technical means alone. Any such attempt suggests that humans are not special but rather are just like machines. Embodied AI also firmly places us within the animal kingdom and uses many insights from evolutionary biology. Any sense of specialness is rejected because there is no empirical evidence that humans are more unique than any other animal. Most of what we think is special in humans can be directly tracked to earlier stages of development in other animals. At the same time, I had learned in theology to understand humans as special, elected by God to be God's partners. Because these two worldviews clashed, I was constantly walking the boundary between them.

What I learned at this time was that doubt is a positive thing. If you enter a dialogue between theology and another discipline and don't think the partner in dialogue has a valid perspective, you might as well not talk at all. Whatever the other party is saying will never convince you or even influence you in any way. If, on the other hand, you throw yourself wholeheartedly into a discussion, you have to question yourself constantly. I often sat in my office at MIT and thought the whole idea of religion was utter rubbish. What I had learned about the human machinery sounded so entirely convincing. These moments of doubt were very painful. But they were also constructive, as I learned the validity of both sides of the argument. I was able to follow Tillich's footsteps in walking the boundary without coming down on one side or the other because I found both sides so fascinating and attractive. I loved the people in both fields and was able to build connections between them.

Could I have found a similar situation in any scientific lab that attempts to understand humans in terms of evolution and functionalism? I think it was the presence of the humanoids in the AI Lab that made the journey so special. Humanoid robots invoke many contrary emotions in most of us. We often perceive them as a threat; we fear they might turn against us. We also resent the possibility that these creatures are as smart as or even smarter than we are, because we feel that we humans are special. But while we fear humanoid robots, we are also attracted to them. We are intrigued by the idea of nonhuman partners inhabiting the earth with us. As a deeply lonely species, we have a strong desire to communicate with beings different from us. Our attempts to communicate with dolphins and chimps and our continuing search for extraterrestrial intelligence demonstrate the depths of this desire. The construction of humanoid robots follows this search for partnership. It can therefore be linked to the Jewish golem tradition in which the construction of humanoid robots is understood as praise of God and as a repetition of God's act of creating us.

Even when I talk about computers, my focus is on humanoid robots and I strictly distinguish between the two. Computers are the machines that sit on your desk or in your lap. They are behind bank machines and at other sites where computing power is needed. Computers can be accessed via keyboard, voice, or touch, and they respond by processing data, running programs, and outputting language.

Robots, on the other hand, are bodies that have specific functions in and interactions with real-world environments through sensors and actuators. Even if their actions are based on computer power, they appear as entities in which computers play only one part. The popularity of fictional humanoid robot characters such as Commander Data (Star Trek) or C-3PO (Star Wars) demonstrate this. Most people are fascinated by robots and endeared by them.

The power of robots lies in their physical presence. They share our space. They move in the world we move in. They interact directly with us, and this allows us to bond with them. Bonding with computers occurs because computers are fascinating machines. Bonding with robots occurs because of their physical reality in this world and our ability to interact with them in physical space.

Among the most powerful science-fiction robots are C-3PO and R2-D2 from Star Wars. R2-D2 is much more darling than C-3PO despite C-3PO's humanoid form and human speech. I think that people prefer R2-D2 because it is so emotional. Even though its beeps are unintelligible to us, we think we understand it because we empathize and bond with it.

This is the same for Cog and Kismet. Cog is more like C-3PO in that it has a humanoid form and invites "grown-up" interaction. Kismet is cute like a baby, fostering an emotional connection-like the one with R2-D2.

This suggests that our bonding mechanisms depend on our own perception of the other and that therefore our ability to bond with them depends much more on emotional settings than on abstract "humanlike" qualities. For the same reason, it is the very emotionality Commander Data from Star Trek displays every time it complains about having no emotions that endears us; an emotionless machine would not constantly raise the issue of its own worth, value, and personhood.

Needless to say, I am very conscious not to assign robots any gender. Even if we, for instance, might think Data is male-after all, it is played by a male actor-I don't want to assign it gender. Humans are accustomed to interacting with creatures that have a gender. Although we find the address "it" offensive, I don't mean it this way. I just want to avoid a "male" impression of the robots.

Cog and Kismet are the first steps in the development of humanoid robots that will be part of our future. But there are many questions about a future with robots. Will these beings be seen as our future partners or as our future enemies? Will we ever perceive robots to be as special as we perceive ourselves to be? Might they make us superfluous in the course of the new, technology-induced evolution?

While I will address many of these questions, I have yet another emotional response to robots. While working with Cog and later Kismet, I have learned more about myself than at any other time. Why would it make me happy if Kismet smiled at me when I knew that it was a programmed reaction? Why would I be disappointed when Cog was ignoring me, even if I knew that-at that time-it could hardly see anything in its periphery?

Our strong emotional reactions toward the robots let the members of my team study themselves. What, after all, do we really know about ourselves as biological creatures, functional systems, and images of God? We decided to explore what modern science could teach us about ourselves. We studied neuroscience to learn more about the brain, and we studied behavioral science and animal studies to learn about mechanisms for common behaviors. We studied evolutionary biology to learn why our bodies are the way they are and what that means for our being in the world. Later on, we concentrated on developmental psychology; adult intelligence is, after all, developed over time, and we thought that if we simulated the learning process of an infant in our robots they would develop a higher degree of intelligence.

This book is the result of my journey at MIT, which started with the monumental handshake between Harvey and Cog. It was this handshake that made me realize that I was not an oddball trying to bring together things that simply do not belong. It convinced me that a dialogue was not just possible but necessary. I finally realized that these robots can serve as thinking tools to explore how we are and how we function in relationships. Surprisingly, the reflection on the robots led me to find some answers to one of the most urgent questions at the beginning of the third millennium. How can we achieve a global community where deeply ingrained cultural differences can, nonetheless, create a community in which different people can live in harmony?

In this book, I explore what robots can teach us about ourselves, our emotions, our ways of thinking and acting in the world. I will talk about what the very wish to build such a creature tells us about Homo sapiens. In the end, we will see that such questions lead to deeper theological and philosophical insights into who we are as thinking machines, as bodies, and as interactive beings and partners of God.

© 2004 Anne Foerst, Ph.D. All rights reserved. No Part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any other information storage and retrieval system, without the written permission of the publisher.

Tags: Religion and Spirituality

About the Author

Dr. Anne Foerst is a former research scientist at the Artificial Intelligence Laboratory at MIT, where she also founded and directed the God and Computers Project. The only robotics theologian in the country, her work has captured much media attention, including coverage in The New York Times, The Boston Globe, and Science. She is currently a visiting professor of theology and computer science at St. Bonaventure University and lives in Olean, New York.