Appeared in the May 10, 2019 print edition of the Wall Street Journal. The title was chosen by the editor.
Science fiction often depicts artificial intelligence as technical minds embodied in humanlike bodies. Think Commander Data of “Star Trek: The Next Generation.” In reality, AI is mindless and usually disembodied. Yet it’s still important, and scientists shouldn’t be the only ones with a say in its future.
The Ethics and Religious Liberty Commission of the Southern Baptist Convention recently laid a marker down with “Artificial Intelligence: An Evangelical Statement of Principles.” The document addresses topics from sex and medicine to accountability and the image of God. The common theme: What does it mean to be human?
It’s encouraging to see religious leaders consider the implications of new technology. Yet as an artificial-intelligence scientist and evangelical Christian, I found the document disappointing.
Its format resembles the Lausanne Covenant of 1974, one of the most important statements in modern evangelicalism. Drafted in an open process, the Lausanne Covenant was adopted by 2,300 delegates from 150 countries. Yet the ERLC document represents only a narrow slice of the global evangelical experience. Most of the signatories are relatively conservative American Protestants, and their seemingly non-sequitur affirmations of just war and traditional marriage reflect that.
The statement only superficially engages the reality of artificial intelligence. It often reads as if the community of AI scientists and ethicists weren’t even consulted. Most signatories are pastors and theologians, and almost none have expertise in artificial intelligence. Recall Proverbs 4:7: “Get understanding before anything else.”
I can help. For more than 15 years, I’ve used artificial intelligence to understand problems that overlap biology, chemistry and medicine. Drawing on a set of generally applicable principles, my colleagues and I use AI to advance scientific knowledge in surprising ways.
At its core, artificial intelligence is little more than a numerical dance, an intricate series of mathematical operations. “Machine learning”—adaptable programs that identify patterns from data—is the type of AI growing in prominence now. My team uses machine learning to understand how and why drugs become toxic, for example, and to determine which kidneys can safely be transplanted.
AI should be guided by a clear ethical framework, but imprecise ethical cautions could cost lives. The document declares “informed consent” to be “requisite.” Yet informed consent can be ethically waived in many cases that are important for academic research. Insisting on informed consent, without acknowledging waivers, has the effect of shackling life-saving scientific work. This omission erects a poorly considered religious barrier to the type of medical research my group does.
The document also states that “moral decision-making” is the exclusive responsibility of humans. Yet artificial intelligence can maneuver a Tesla. In an accident, the car may need to make moral decisions. Risking the safety of its passengers, should the vehicle dangerously swerve to avoid a pedestrian? The document seems to oppose artificial intelligence where it might delegate moral decisions like this—but it’s not clear. This risks unnecessary prohibitions of life-saving technology.
Speaking of artificial intelligence in the future, the document notes that “God alone has the power to create life.” This phrase appears in traditional theology as an affirmation of God’s providence and authority. Of course it doesn’t prohibit creating life through reproduction. Nor does it proscribe scientific work like creating new viruses in a lab. Citing “God alone,” nonetheless, the document seems to declare artificial minds either impossible or immoral. Why not encourage scientific inquiry?
A “person” like Commander Data has yet to emerge from the electronics of a computer. At the same time, the human mind is somehow entwined in the electronic fluctuations of neurons in our brains. Science always calls for humility, and it’s true that AI already has shown surprising linguistic, artistic and social abilities. Yet none of these feats even remotely approach demonstration of a humanlike mind.
Grand questions loom. Can a computer house a mind? How would scientists and engineers construct a computational mind? How would they know if they succeeded? We can’t know for sure, but these questions welcome all of us, including theologians. Rather than offering a far- reaching statement of religious convictions, it would be better to start with a list of questions.