Article Artifi cial Intelligence: Discerning a Christian Response Derek C. Schuurman Derek C. Schuurman Artifi cial Intelligence (AI) techniques employing deep learning have recently achieved remarkable strides in tackling diffi cult problems and spurring applications in many new areas. Responses to these developments have ranged from existential fear to unbridled optimism. This technology opens up a plethora of ethical considerations and ontologi- cal questions about what it means to be human. The approach one takes to questions arising in AI is largely shaped by our philosophical presuppositions and our world- view. This article sketches some of the implications and important questions that arise surrounding AI. The article concludes by urging Christians to join this conversation, bringing insights from scripture and from Christian philosophy and theology to inform a responsible approach that contributes to the common good. he movie Wall-E is an entertain- For many decades, there have been many ing tale of a dystopian future of optimistic predictions about the capabili- Trobots, automation, and human- ties of Artifi cial Intelligence (AI) which ity. A polluted earth is left abandoned have consistently fallen short of expecta- except for robots like the charming title tions. In 1958 Frank Rosenblatt pioneered character Wall-E, who are left to clean up modeling neurons using simple networks the mess. Humans have fl ed the planet, called “perceptrons” which could be coddled aboard a massive ark-like space- trained to classify data. Later, the pioneer- ship where automated systems take care ing AI researchers Marvin Minsky and of their every need. It is striking that the Seymour Papert published an infl uential most human-like characters in the movie book titled Perceptrons which identifi ed are the two main robot characters while challenges with single-layer perceptrons the human characters are portrayed as and expressed skepticism about multi- obese, feeble, and passive, shuttled about layer perceptrons. They wrote, in reclining chairs, and consuming bev- Perceptrons have been widely erages while perpetually entertained publicized as “pattern recognition” by personal screens. At the climax of or “learning machines” and as the movie, the ship’s captain valiantly such have been discussed in a large struggles to stand and, unaccustomed to number of books, journal articles, and walking, waddles over to the main con- voluminous “reports.” Most of this trol panel to wrestle control back from the writing ... is without scientifi c value.1 automated ship. The tension in this cli- As a result, work in this area diminished mactic moment is driven by one question: greatly through the 1970s, during an era will humanity take back control from technology? Derek C. Schuurman worked as an electrical engineer for a number of years before returning to school to complete a PhD in the area of robotics Please note: A draft of this article was originally posted online in January 2018 as an invitational and computer vision. He is now a professor of computer science at Calvin essay with a Call for Papers for a special issue on College in Grand Rapids, Michigan, where he currently holds the William Artifi ciaI Intelligence. The two articles which follow Spoelhof Teacher-Scholar-in-Residence chair. He is the author of the book were subse quently submitted and reviewed in Shaping a Digital World: Faith, Culture and Computer Technology response to this invitational essay. (InterVarsity Press). Volume 71, Number 2, June 2019 75 Article Artifi cial Intelligence: Discerning a Christian Response sometimes referred to as an “AI winter.” However, pages). These books include titles such as Technology interest in multilayer perceptrons was reignited in vs. Humanity: The Coming Clash between Man and the mid-1980s after various breakthrough papers Machine; In Our Own Image: Savior or Destroyer? The were published demonstrating how they could History and Future of Artifi cial Intelligence; and The be made effective by employing specialized train- Glass Cage: Automation and Us. Some of these books ing algorithms.2 These techniques have since been take an optimistic stance, some are more circum- further refi ned, and, combined with advances in spect, while others paint a darker picture. computing power, have led to so-called “deep-learn- ing” methods.3 Some have suggested that the advance of technol- ogy and AI will eventually solve all our problems. Deep learning uses many layers of perceptrons The term technicism is a word that has been coined which can be trained using special techniques such to refer to the faith in technology as savior or rescuer as backpropagation or gradient descent. Deep learn- of the human condition.7 A recent book titled Infi nite ing is an approach to machine learning, a fi eld which Progress includes the subtitle: “How the Internet and involves training computers to “learn” patterns Technology Will End Ignorance, Disease, Poverty, without being explicitly programmed for those pat- Hunger, and War.”8 This is essentially a form of terns. The training process will typically employ idolatry, replacing a trust in the Creator with tech- a labeled set of example training data in a process nology. In fact, this trust in technology becomes called “supervised learning.” Alternately, training explicit in the case of the “Way of the Future,” a reli- can also be performed using a set of unlabeled input gious group founded by Anthony Levandowski, a data which is then processed to uncover patterns and former Google and Uber engineer who is working to structures. That process is referred to as “unsuper- “develop and promote the realization of a Godhead vised learning.” based on Artifi cial Intelligence” and that “through understanding and worship of the Godhead, [to] AI techniques employing deep learning have contribute to the betterment of society.”9 The trans- recently achieved remarkable strides in tackling humanist Zoltan Istvan suggests that this new AI more diffi cult problems. A research team at Google deity “will actually exist and hopefully will do things demonstrated these techniques by developing a sys- for us.”10 These sentiments are explicit examples of tem that was trained to play the game Go by playing an observation made by the writer David Noble that games against itself, eventually surpassing even the “the technological enterprise has been and remains best human players.4 Google has recently released suffused with religious belief.”11 its machine learning library, TensorFlow, under an open source license, spurring applications in many Everyone has a worldview which informs a set of new areas.5 These tools are not just solving puz- beliefs that shape our conception of reality. Nicholas zles in the laboratory. They are now being directed Wolterstorff suggests it is these “control beliefs” that toward a plethora of diffi cult practical problems enable us to commit to a particular theory.12 These that traditionally have been beyond the capabili- beliefs are also active in our technical work, includ- ties of prior AI systems. For instance, these systems ing the theories related to research in AI, whether are showing great promise in diagnosing certain explicitly stated or not. diseases and analyzing medical images, even out- Some engineers and computer scientists believe that performing human doctors in some tasks.6 AI is also technology will even solve the problem of death. making advances in diverse areas such as legal work, According to David Pearce, co-founder of an organi- image recognition, and language translation. The rise zation called Humanity+: of autonomous vehicles is another emerging area in which deep learning has made remarkable progress. If we want to live in paradise, we will have to engineer it ourselves. If we want eternal life, then As a book review editor for PSCF on topics relat- we’ll need to rewrite our bug-ridden genetic code ing to technology, I have been astounded at the and become god-like … only hi-tech solutions can sheer number of books that have been released in ever eradicate suffering from the living world.13 recent years about issues surrounding AI and robot- Ray Kurzweil, an accomplished computer scientist ics (several of which have been reviewed in these and author of The Age of Spiritual Machines, has sug- 76 Perspectives on Science and Christian Faith Derek C. Schuurman gested that within the present century we will be able While these threats may seem far-fetched, the more to upload our brain into a computer and live forever, immediate concern is the loss of jobs due to AI, free from the limitations of our mortal bodies. This robots, and automation. In the early 2000s, I was idea has been coined the “rapture of the geeks,” and doing my graduate studies in the area of computer Kurzweil writes, “We don’t always need real bodies. vision. At the time, I recall thinking that self-driving If we happen to be in a virtual environment, then a cars were unlikely to be feasible due to the challenges virtual body will do just fi ne.”14 of real-time vision systems in unstructured environ- ments. However, within a short decade, autonomous David F. Noble observes, vehicles were successfully demonstrated. In the near Artifi cial Intelligence advocates wax eloquent term, autonomous vehicles are likely to disrupt the about the possibilities of machine-based immor- labor market, potentially displacing millions of jobs tality and resurrection, and their disciples, the in driving professions. architects of virtual reality and cyberspace, exult in their expectation of God-like omnipresence and One paper published by researchers from the disembodied perfection.15 University of Oxford predicts that 47 percent of U.S. Psalm 115 states that the makers of idols will become jobs are at risk of being replaced by AI technolo- 17 like them, and in the case of the “rapture of the gies and computerization.