When the Vatican Talks About Artificial Intelligence, Pay Attention

A Major Note on Artificial Intelligence Has Dropped

Three, maybe four, things to pay attention to in one of the first major ecclesial treatments of artificial intelligence.

The Commentary on Artificial Landscape? Meet the Vatican

Theological commentary on artificial intelligence is no recent event. Presbyterians, Methodists, and evangelicals are among the many who have offered commentary in various forms and in innumerable outlets, but three things are worth flagging here, broadly speaking:

1. Most popular commentary on AI speaks with either knowledge of the material claims of AI, or the theological stakes, but not both. I am not among those who can speak with any competency about the nuances of AI, though I have colleagues in computer science that do ground-breaking work on it. James in particular, noted in the story, has a Masters in Divinity, after which he obtained his grad work in computer science. But he’s one of a handful of folks with that kind of competency. Other thinktanks, like this one, bring together folks with different competencies do interdisciplinary work, but their concerns are more interfaith, and thus (to my mind) wind up being pretty milquetoast in conclusions. Vatican statements on social questions, by contrast, speak with careful competency on the material questions, but also deep engagement with theological wisdom and authorities. It’s the Vatican after all.

2. Most popular commentary speaks for the individual representatives, but not whole traditions. This, again, sets Vatican statements apart in most cases: when the Vatican issues guidance on a question, it’s for all of the faithful. The National Council of Catholic Bishops can differ with interpretative points, or the German bishops can go rogue on points, but Vatican documents propose to speak normatively on a Christian vision of the question at hand. The documents will typically point back to other documents which more fully exposit certain subpoints, inviting the reader into a stream of thinking and reflection, as opposed to a series of fragmented one-off statements.

3. Vatican documents tend to not get out over their skiis. Rather than getting tied up in policy, Vatican documents offer guiding points for action and reflection. Since the 1890s, the tradition of writings known collectively as Catholic Social Teaching has threaded this needle on everything from migration to abortion to labor concerns to war to cremation. Typically, you’ll find a few points of normative instruction for practice, but frequently, these are guardrails—this point and no further—leaving room for prudential deliberation. The Internet tends to be the worst on this: offering overconfident expositions instead of modest direction.

With that in mind, three-and-a-half points from yesterday’s “Note on the Relationship Between Artificial Intelligence and Human Intelligence”. You can read the full document in English here.

This is what AI generates if you ask it to generate a picture of “artificial intelligence.” The near devotional image, almost Marian image, should tip us off that AI—in engaging with humans—is responding to a desire for integration into human life at profound levels.

Three-And-A-Half Points Of Note

1. The Focus on What Constitutes Intelligence

After offering an overview of the development and broad aims of AI, the following clarifier is made:

12. AI’s advanced features give it sophisticated abilities to perform tasks, but not the ability to think.[12] This distinction is crucially important, as the way “intelligence” is defined inevitably shapes how we understand the relationship between human thought and this technology.[13]

By drawing a distinction between the computational nature of AI and the nature of thinking, the document puts in clear writing what is and is not at stake. While AI is increasingly capable of passing the Turing Test, this is not the same as thinking:

This functional perspective is exemplified by the “Turing Test,” which considers a machine “intelligent” if a person cannot distinguish its behavior from that of a human.[11] However, in this context, the term “behavior” refers only to the performance of specific intellectual tasks; it does not account for the full breadth of human experience, which includes abstraction, emotions, creativity, and the aesthetic, moral, and religious sensibilities. Nor does it encompass the full range of expressions characteristic of the human mind. Instead, in the case of AI, the “intelligence” of a system is evaluated methodologically, but also reductively, based on its ability to produce appropriate responses—in this case, those associated with the human intellect—regardless of how those responses are generated.

Put differently, if we presume that “thinking” is a matter of recitation of facts or computational ability, we as a culture have adopted a remarkably inhuman vision of what thought is, and what intelligence consists of.

2. An Exposition of Human Intelligence and Its Aims

There have been a number of significant Vatican statements on the nature of reason, including (but not limited to) Fides et Ratio, and here we find a distillation of teaching on the integral nature of thought. Thought, as exposited here, is different than computation, for the human mind is composed of intellect, passion, and most significantly, an orientation by God toward knowing the ends for which minds and intelligence exist. Intelligence occurs as we proceed through the world, as those kinds of creatures whose bodies and minds are inseparable. In the same way that we learn discipleship not as an intellectual exercise, but as a practice of the body, so we always think as those kinds of creatures with passions, pasts, and futures.

Thinking is an integral activity, the document contends:

26. In this context, human intelligence becomes more clearly understood as a faculty that forms an integral part of how the whole person engages with reality. Authentic engagement requires embracing the full scope of one’s being: spiritual, cognitive, embodied, and relational.

27. This engagement with reality unfolds in various ways, as each person, in his or her multifaceted individuality[54], seeks to understand the world, relate to others, solve problems, express creativity, and pursue integral well-being through the harmonious interplay of the various dimensions of the person’s intelligence.[55] This involves logical and linguistic abilities but can also encompass other modes of interacting with reality. Consider the work of an artisan, who “must know how to discern, in inert matter, a particular form that others cannot recognize”[56] and bring it forth through insight and practical skill. Indigenous peoples who live close to the earth often possess a profound sense of nature and its cycles.[57]

Similarly, a friend who knows the right word to say or a person adept at managing human relationships exemplifies an intelligence that is “the fruit of self-examination, dialogue and generous encounter between persons.”[58] As Pope Francis observes, “in this age of artificial intelligence, we cannot forget that poetry and love are necessary to save our humanity.”[59]

3. The Limits of AI for Moral Reflection

There is a growing trend, amplified in the “Effective Altruist” movement, which is comfortable using computational thinking to resolve difficult moral questions of scale: What does it look like to preserve a world in which food is available? How do we deal with diminished resources? How should we live amidst climate change?

These kinds of macro-questions are important ones, but also can be resolved at scale primarily by slimming humans down into interchangeable parts. Utilitarian thought trades on defining happiness, and maximal happiness, as the desirable goal, and then building out ways of achieving it from there. But happiness isn’t the kind of thing which is a one-sized-fits-all goal, nor is happiness, on its face, a great goal to aim for. Some forms of happiness, after all, are manifestly bad for people, and some are morally egregious forms of happiness.

This last part exposes the limits of AI for resolving difficult moral problems at scale, and the document concurs, in a limited way:

41. At the same time, it is not only the ends that are ethically significant but also the means employed to achieve them. Additionally, the overall vision and understanding of the human person embedded within these systems are important to consider as well. Technological products reflect the worldview of their developers, owners, users, and regulators,[86] and have the power to “shape the world and engage consciences on the level of values.”[87] On a societal level, some technological developments could also reinforce relationships and power dynamics that are inconsistent with a proper understanding of the human person and society.

42. Therefore, the ends and the means used in a given application of AI, as well as the overall vision it incorporates, must all be evaluated to ensure they respect human dignity and promote the common good.[88] As Pope Francis has stated, “the intrinsic dignity of every man and every woman” must be “the key criterion in evaluating emerging technologies; these will prove ethically sound to the extent that they help respect that dignity and increase its expression at every level of human life,”[89] including in the social and economic spheres. In this sense, human intelligence plays a crucial role not only in designing and producing technology but also in directing its use in line with the authentic good of the human person.[90]

The language here of “common good” and “human dignity” is left open. It’s not the Vatican doesn’t have specifics in mind, but it points you to the footnotes where these more specific questions have been reasoned about in much more detail, and it presumes deep Scriptural engagments as well in those prior documents.¹

The emphasis here is on keeping humans intimately involved with the workings of these computations, in that AI—unless informed in the kinds of things it should pay attention to—won’t know to treat humans as the kinds of complex, culture-informed, intricate creatures they are. The results will be either to propose low-hanging fruit which doesn’t touch the depths of the issues, or to propose inhuman solutions in which humans are the problem to be solved.

3a. AI as A Cultural Problem

The document concludes with a long series of brief engagements with education, warfare, deepfakes, ecology, and more. But what binds these disparate comments together is a concern for how AI is embedded within so many of our cultural components already. It’s not enough to say “No AI”, insofar as commerce, governmental decisions, and ordinary technologies already make use of these. The first step toward loosening their determinist grip on cultures, though, is to be aware of what they are doing.

I would encourage you to read the last half of the document with care, noting that what binds these together is a persistent attention to the complexity of the human person. If AI is a computational reality, then we cannot simply offload engagements with the world to forms of management and hope for the best. It requires the difficult work of attention, and more specifically, attending to people as guided by virtue. It’s easy to pay attention to people reductively, as categories, classes, or headcounts. But it’s harder to pay attention with the complexity that the document presumes is necessary to pay attention to actual human beings.

Take and read.

1. This is a good practice for public writers to adopt: just refer the reader to where you’ve dealt with that at length. If you’ve never dealt with it at length, that’s a good sign that maybe you’ve gone too quickly and haven’t dealt with foundational questions first.