Marius Dorobantu, Author at AI Theology

Vulnerable like God: Perfect Machines and Imperfect Humans

This four-part series started with the suggestion that AI can be of real help to theologians, in their attempt to better understand what makes humans distinctive and in the image of God. We have since noted how different machine intelligence is from human intelligence, and how alien-like an intelligent robot could be ‘on the inside’, in spite of its humanlike outward behavior.

For theological anthropology, the main takeaway is that intelligence – understood as rationality and problem-solving – is not the defining feature of human nature. We’ve long been the most intelligent and capable creature in town, but that might soon change, with the emergence of AI. What makes us special and in the image of God is thus not some intellectual capacity (in theology, this is known as the substantive interpretation), nor something that we can do on God’s behalf (the functional interpretation), because AI could soon surpass us in both respects.

The interpretation of the imago Dei that seems to account best for the scenario of human-level AI is the relational one. According to it, the image of God is our special I-Thou relationship with God, the fact that we can be an authentic Thou, capable to receive God’s love and respond back. We exist only because God calls us into existence. Our relationship with God is therefore the deepest foundation of our ontology. Furthermore, we are deeply relational beings. Our growth and fulfillment can only be realized in authentic personal relationships with other human beings and with God.

AI and Authentic Relationality

It is not surprising that the key to human distinctiveness is profoundly relational. Alan Turing tapped right into this intuition when he designed his eponymous test for AI. Turing’s test is, in fact, a measurement of AI’s ability to relate like us. Unsurprisingly, the most advanced AIs still struggle when it comes to simulating relationships, and none has yet passed the Turing test.

But even if a program will someday convincingly relate to humans, will that be an authentic relationship? We’ve already seen that human-level AI will be anything but humanlike ‘on the inside.’ Intelligent robots might become capable to speak and act like us, but they will be completely different from us in terms of their internal motivation or meaning systems. What kind of relationship could there be between us and them, when we’d have so little in common?

We long for other humans precisely because we are not self-sufficient. Hence, we seek others precisely because we want to discover them and our own selves through relationships. We fall in love because we are not completely rational. Human-level AI will be the opposite of that: self-sufficient, perfectly rational, and with a quasi-complete knowledge of itself.

The Quirkiness of Human intelligence

Our limitations are instrumental for the kind of relationships that we have with each other. An argument can thus be made that a significant degree of cognitive and physical vulnerability is required for authentic relationality to be possible. There can be no authentic relationship without the two parts intentionally making themselves vulnerable to each other, opening to one another outside any transactional logic.

Photo by Duy Pham on Unsplash

A hyper-rational being would likely have very serious difficulties to engage fully in relationships and make oneself totally vulnerable to the loved other. It surely does not sound very smart.

Nevertheless, we, humans, do this tirelessly and often at high costs, exactly, perhaps, because we are not that intelligent and goal oriented as AI. Although that appears to be illogic, it is such experiences that give meaning and fulfillment to our lives.

From an evolutionary perspective, it is puzzling that our species evolved to be this way. Evolution promotes organisms that are better at adapting to the challenges of their environment, thus at solving practical survival and reproduction problems. It is therefore unsurprising that intelligence-as-problem-solving is a common feature of evolved organisms, and this is precisely the direction in which AI seems to develop.

If vulnerability is so important for the image of God as relationship, does this imply that God too is vulnerable?

What is strange in the vast space of possible intelligences is our quirky type of intelligence, one heavily optimized for relationship, marked by a bizarre thirst for meaning, and plagued by a surprising degree of irrationality. In the previous post I called out the strangeness of strong AI, but it is we who seem to be strange ones. However, it is specifically this kind of intellectual imperfection, or vulnerability, that enables us to dwell in the sort of Goldilocks of intelligence where personal relationships and the image of God are possible.

Vulnerability, God, Humans and Robots

If vulnerability is so important for the image of God as relationship, does this imply that God too is vulnerable? Indeed, that seems to be the conclusion, and it’s not surprising at all, especially when we think of Christ. Through God’s incarnation, suffering, and voluntary death, we have been revealed a deeply vulnerable side of the divine. God is not an indifferent creator of the world, nor a dispassionate almighty, all-intelligent ruler. God cares deeply for creation, to the extent of committing to the supreme self-sacrifice to redeem it (Jn. 3: 16).

This means that we are most like God not when we are at our smartest or strongest, but when we engage in this kind of hyper-empathetic, though not entirely logical, behavior.

Compared to AI, we might look stupid, irrational, and outdated, but it is paradoxically due to these limitations that we are able to cultivate our divine likeness through loving, authentic, personal relationships. If looking at AI teaches theologians anything, it is that our limitations are just as important as our capabilities. We are vulnerable, just as our God has revealed to be vulnerable. Being like God does not necessarily mean being more intelligent, especially when intelligence is seen as rationality or problem solving

Christ – whether considered historically or symbolically – shows that what we value most about human nature are traits like empathy, meekness, and forgiveness, which are eminently relational qualities. Behind such qualities are ways of thinking rooted more in the irrational than in the rational parts of our minds. We should then wholeheartedly join the apostle Paul in “boast[ing] all the more gladly about [our] weaknesses […] for when [we] are weak, then [we] are strong” (2 Cor. 12: 9-10).

Human-level, but not Humanlike: The Strangeness of Strong AI

The emergence of AI opens up exciting new avenues of thought, promising to add some clarity to our understanding of intelligence and of the relation between intelligence and consciousness. For Christian anthropology, observing which aspects of human cognition are easily replicated in machines can be of particular help in refining the theological definition of human distinctiveness and the image of God.

However, by far the most theologically exciting scenario is the possibility of human-level AI, or artificial general intelligence (AGI), the Holy Grail of AI research. AGI would be capable to convincingly replicate human behavior. It could, in principle, pass as human, if it chose to. This is precisely how the Turing Test is designed to work. But how humanlike would a human-level AI really be?

Computer programs have already become capable of impressive things, which, when done by humans, require some of our ‘highest’ forms of intelligence. However, the way AI approaches such tasks is very non-humanlike, as explained in the previous post. If the current paradigm continues its march towards human-level intelligence, what could we expect AGI to be like? What kind of creature might such an intelligent robot be? How humanlike would it be? The short answer is, not much, or even not at all.

The Problem of Consciousness

Philosophically, there is a huge difference between what John Searle calls ‘strong’ and ‘weak’ AI. While strong AI would be an emulation of intelligence, weak AI would be a mere simulation. The two would be virtually indistinguishable on the ‘outside,’ but very different ‘on the inside.’ Strong AI would be someone, a thinking entity, endowed with conscience, while weak AI would be a something, a clockwork machine completely empty on the inside.

It is still too early to know whether AGI will be strong or weak, because we currently lack a good theory of how consciousness arises from inert matter. In philosophy, this is known as “the hard problem of consciousness.” But if current AI applications are any indication, weak AGI is a much more likely scenario than strong AGI. So far, AI has made significant progress in problem-solving, but it has made zero progress in developing any form of consciousness or ‘inside-out-ness.’

Even if AGI does somehow become strong AI (how could we even tell?), there are good reasons to believe that it would be a very alien type of intelligence.

What makes human life enjoyable is arguably related to the chunk of our mind that is not completely rational.

Photo by Maximalfocus on Unsplash

The Mystery of Being Human

As John McCarthy – one of the founding fathers of AI – argues, an AGI would have complete access to its internal states and algorithms. Just think about how weird that is! Humans have a very limited knowledge of what happens ‘on their inside.’ We only see the tip of the iceberg, because only a tiny fraction of our internal processes enter our stream of consciousness. Most information remains unconscious, and that is crucial for how we perceive, feel, and act.

Most of the time we have no idea why we do the things we do, even though we might fabricate compelling, post-factum rationalizations of our behavior. But would we really want to know such things and always act in a perfectly rational manner? Or, even better, would we be friends or lovers with such a hyper-rational person? I doubt.

Part of what makes us what we are and what makes human life enjoyable is arguably related to the chunk of our mind that is not completely rational. Our whole lives are journeys of self-discovery, and with each experience and relationship we get a better understanding of who we are. That is largely what motivates us to reach beyond our own self and do stuff. Just think of how much of human art is driven precisely by a longing to touch deeper truths about oneself, which are not easily accessible otherwise.

Strong AI could be the opposite of that. Robots might understand their own algorithms much better than we do, without any need to discover anything further. They might be able to communicate such information directly as raw data, without needing the approximation/encryption of metaphors. As Ian McEwan’s fictitious robot character emphatically declares, most human literature would be completely redundant for such creatures.

The Uncanny Worldview of an Intelligent Robot

Intelligent robots would likely have a very different perception of the world. With access to Bluetooth and WiFi, they would be able to ‘smell’ other connected devices and develop a sort of ‘sixth sense’ of knowing when a particular person is approaching merely from their Bluetooth signature. As roboticist Rodney Brooks shows, robots will soon be able to measure one’s breathing and heart rate without any biometric sensor, simply by analyzing how a person’s physical presence slightly changes the behavior of WiFi signals.

The technology for this already exists, and it could enable the robot to have access to a totally different kind of information about the humans around, such as their emotional state or health. Similar technologies of detecting changes in the radio field could allow the robots to do something akin to echolocation and know if they are surrounded by wood, stone, or metal. Just imagine how alien a creature endowed with such senses would be!

AGI might also perceive time very differently from us because they would think much faster. The ‘wetware’ of our biological brains constrains the speed at which electrical signals can travel. Electronic brains, however, could enable speeds closer to the ultimate physical limit, the speed of light. Minds running on such faster hardware would also think proportionally faster, making their experience of the passage of time proportionally slower.

If AGI would think ‘only’ ten thousand times faster than humans, a conservative estimation, they would inhabit a completely different world. It is difficult to imagine how such creatures might regard humans, but futurist James Lovelock chillingly estimates that “the experience of watching your garden grow gives you some idea of how future AI systems will feel when observing human life.”

The way AGI is depicted in sci-fi (e.g. Terminator, Ex Machina, or Westworld) might rightly give us existential shivers. But if the predictions above are anywhere near right, then AGI might turn out to be weirder than our wildest sci-fi dreams. AI might reach human-level, but it would most likely be radically non-humanlike.

Is this good or bad news for theological anthropology? How would the emergence of such an alien type of affect our understanding of humanity and its imago Dei status? The next post, the last one in this four-part series, wrestles head-on with this question.

How Does AI Compare with Human Intelligence? A Critical Look

In the previous post I argued that AI can be of tremendous help in our theological attempt to better understand what makes humans distinctive and in the image of God. But before jumping to theological conclusions, it is worth spending some time trying to understand what kind of intelligence machines are currently developing, and how much similarity is there between human and artificial intelligence.Image by Gordon Johnson from Pixabay

The short answer is, not much. The current game in AI seems to be the following: try to replicate human capabilities as well as possible, regardless of how you do it. As long as an AI program produces the desired output, it does not matter how humanlike its methods are. The end result is much more important than what goes on ‘on the inside,’ even more so in an industry driven by enormous financial stakes.

Good Old Fashioned AI

This approach was already at work in first wave of AI, also known as symbolic AI or GOFAI (good old-fashioned AI). Starting with the 1950s, the AI pioneers struggled to replicate our ability to do math and play chess, considered the epitome of human intelligence, without any real concern for how such results were achieved. They simply assumed that this must be how the human mind operates at the most fundamental level, through the logical manipulation of a finite number of symbols.

GOFAI ultimately managed to reach human-level in chess. In 1996, an IBM program defeated the human world-champion, Gary Kasparov, but it did it via brute force, by simply calculating millions of variations in advance. That is obviously not how humans play chess.

Although GOFAI worked well for ‘high’ cognitive tasks, it was completely incompetent in more ‘mundane’ tasks, such as vision or kinesthetic coordination. As roboticist Hans Moravec famously observed, it is paradoxically easier to replicate the higher functions of human cognition than to endow a machine with the perceptive and mobility skills of a one-year-old. What this means is that symbolic thinking is not how human intelligence really works.

The Advent of Machine Learning

What replaced symbolic AI since roughly the turn of the millennium is the approach known as machine learning (ML). One subset of ML that has proved wildly successful is deep learning, which uses layers of artificial neural networks. Loosely inspired by the brain’s anatomy, this algorithm aims to be a better approximation of human cognition. Unlike previous AI versions, it is not instructed on how to think. Instead, these programs are being fed huge sets of selected data, in order to develop their own rules for how the data should be interpreted.

For example, instead of teaching an ML algorithm that a cat is a furry mammal with four paws, pointed ears, and so forth, the program is trained on hundreds of thousands of pictures of cats and non-cats, by being ‘rewarded’ or ‘punished’ every time it makes a guess about what’s in the picture. After extensive training, some neural pathways become strengthened, while others are weakened or discarded. The end result is that the algorithm does learn to recognize cats. The flip side, however, is that its human programmers no longer necessarily understand how the conclusions are reached. It is a sort of mathematical magic.

ML algorithms of this kind are behind the impressive successes of contemporary AI. They can recognize objects and faces, spot cancer better than human pathologists, translate text instantly from one language to another, produce coherent prose, or simply converse with us as smart assistants. Does this mean that AI is finally starting to think like us? Not really.

When machines fail, they fail badly, and for different reasons than us.

Even when machines manage to achieve human or super-human level in certain cognitive tasks, they do it in a very different fashion. Humans don’t need millions of examples to learn something, they sometimes do very fine with at as little as one example. Humans can also usually provide explanations for their conclusions, whereas ML programs are often these ‘black boxes’ that are too complex to interrogate.

More importantly, the notion of common sense is completely lacking in AI algorithms. Even when their average performance is better than that of human experts, the few mistakes that they do make reveal a very disturbing lack of understanding from their part. Images that are intentionally perturbed so slightly that the adjustment is imperceptible to humans can still cause algorithms to misclassify them completely. It has been shown, for example, that sticking minuscule white stickers, almost imperceptible to the human eye, on a Stop sign on the road causes the AI algorithms used in self-driving vehicles to misclassify it as a Speed Limit 45 sign. When machines fail, they fail badly, and for different reasons than us.

Machine Learning vs Human Intelligence

Perhaps the most important difference between artificial and human intelligence is the former’s complete lack of any form of consciousness. In the words of philosophers Thomas Nagel and David Chalmers, “it feels like something” to be a human or a bat, although it is very difficult to pinpoint exactly what that feeling is and how it arises. However, we can intuitively say that very likely it doesn’t feel like anything to be a computer program or a robot, or at least not yet. So far, AI has made significant progress in problem-solving, but it has made zero progress in developing any form of consciousness or ‘inside-out-ness.’

Current AI is therefore very different from human intelligence. Although we might notice a growing functional overlap between the two, they differ strikingly in terms of structure, methodology, and some might even say ontology. Artificial and human intelligence might be capable of similar things, but that does not make them similar phenomena. Machines have in many respects already reached human level, but in a very non-humanlike fashion.

For Christian anthropology, such observations are particularly important, because they can inform how we think of the developments in AI and how we understand our own distinctiveness as intelligent beings, created in the image of God. In the next post, we look into the future, imagining what kind of creature an intelligent robot might be, and how humanlike we might expect human-level AI to become.

Artificial Intelligence: The Disguised Friend of Christian Anthropology

AI is making one significant leap after the other. Computer programs can nowadays convincingly converse with us, generate plausible prose, diagnose disease better than human experts, and totally trash us in strategy games like Go and chess, once considered the epitome of human intelligence. Could they one day reach human-level intelligence? It would be extremely unwise to discount such a possibility without very good reasons. Time, after all, is on AI’s side, and the kind of things that machines are capable of today used to be seen as quasi-impossible just a generation ago.

How could we possibly speak of human distinctiveness when robots become indistinguishable from us?

The scenario of human-level AI, also known as artificial general intelligence (AGI), would be a game-changer for every aspect of human life and society, but it would raise particularly difficult questions for theological anthropology. Since the dawn of the Judeo-Christian tradition, humans have perceived themselves as a creature unlike any other. The very first chapter of the Bible tells us that we are special, because only we of all creatures are created in the image of God (imago Dei). However, the Copernican revolution showed us that we are not the center of the universe (not literally, at least), and the Darwinian revolution revealed that we are not ontologically different from non-human animals. AGI is set to deliver the final blow, by conquering the last bastion of our distinctiveness: our intelligence.

By definition, AGI would be capable of doing anything that a standard human can do, at a similar or superior level. How could we possibly speak of human distinctiveness when robots become indistinguishable from us? Christian anthropology would surely be doomed, right? Well, not really, actually quite the contrary. Instead of rendering us irrelevant and ordinary, AI could in fact represent an unexpected opportunity to better understand ourselves and what makes us in God’s image.

Science’s Contribution to the Imago Dei

To explain why, it is useful to step back a little and acknowledge how much the imago Dei theology has benefitted historically from an honest engagement with the science of its time. Based solely on the biblical text, it is impossible to decide what the image of God is supposed to mean exactly. The creation story in Genesis 1 tells us that only humans are created in the image and likeness of God, but little else about what the image means. The New Testament does not add much, except for affirming that Jesus Christ is the perfect image. Ever since Patristic times, Christian anthropology has constantly wrestled with how to define the imago Dei, without much success or consensus.

The obvious way to tackle the question of our distinctiveness is to examine what differentiates us from the animals, the only others with which we can meaningfully compare ourselves. For the most part of Christian history, this difference has been located in our intellectual capacities, an approach heavily influenced by Aristotelian philosophy, which defined the human as the rational animal. But then came Darwin and showed us that we are not as different from the animals as we thought we were.

Theologian Aubrey Moore famously said that Darwin “under the guise of a foe, did the work of a friend” for Christianity.

Furthermore, the following century and a half of ethology and evolutionary science revealed that our cognitive capacities are not bestowed upon us from above. Instead, they are rooted deep within our evolutionary history, and most of them are shared with at least some of the animals. If there is no such thing as a uniquely human capacity, then surely we were wrong all along to regard ourselves as distinctive, right?

Not quite. Theologian Aubrey Moore famously said that Darwin “under the guise of a foe, did the work of a friend” for Christianity. Confronted with the findings of evolutionary science, theologians were forced to abandon the outdated Aristotelian model of human distinctiveness and look for more creative ways to define the image of God. Instead of equating the image with a capacity that humans have, post-Darwinian theology regards the imago Dei in terms of something we are called to do or to be.

Defining the Imago Dei

Some theologians interpret the image functionally, as our election to represent God in the universe and exercise stewardship over creation. Others go for a relational interpretation, defining the image through the prism of the covenantal ‘I-Thou’ relationship that we are called to have with God, which is the fundament of human existence. To be in the image of God is to be in a personal, authentic relationship with God and with other human beings. Finally, there are others who interpret the imago Dei eschatologically, as a special destiny for human beings, a sort of gravitational pull that directs us toward existential fulfilment in the fellowship with God, in the eschaton. Which of these interpretations is the best? Well, hard to say. Without going into detail, let’s just say that there are good theological arguments for each of them.

If purely theological debate does not produce clear answers, we might then try to compare ourselves with the animals. This, though, does not lead us too far either. Although ‘technically’ we are not very different from the animals and we share with them similar bodily and cognitive structures, in practical terms the difference is huge. Our mental lives, our societies and our achievements are so radically different than theirs, that it is actually impossible to pinpoint just one dimension that represents the decisive difference. Animals are simply no match for us. This is good news for human distinctiveness, but it also means that we might be stuck in a never-ending theological debate on how to interpret the image of God, with so many options on our hand.

How Can AI help Define Who We Are?

This is where the emergence of human-level AI can be a game-changer. For the first time, we would be faced with the possibility of an equal or superior other, one that could potentially (out)match us in everything, from our intellectual capacities, to what we can do in the world, our relational abilities, or the complexity of our mental lives. Instead of agonizing about AI replacing us or rendering us irrelevant, we could relish the opportunity to better understand our distinctiveness through the insights brought about by the emergence of this new other.

The hypothesis of AGI might present theologians with an extraordinary opportunity to narrow down their definitions of human distinctiveness and the image of God. Looking at what would differentiate us from human-level AI, if indeed anything at all, may provide just the right amount of conceptual constraint needed for a better definition of the imago Dei. In this respect, our encounter with AI might prove to be our best shot at comparing ourselves with a different type of intelligence, apart from maybe the possibility of ever finding extra-terrestrial intelligence in the universe.

Dr. Marius Dorobantu is a research associate in science & theology at VU Univ. Amsterdam (NL). His PhD thesis (2020, Univ. of Strasbourg, FR) analysed the potential challenges of human-level AI for theological anthropology. The project he’s currently working on, funded by the Templeton WCF within the “Diverse Intelligences” initiative, is entitled ‘Understanding spiritual intelligence: psychological, theological, and computational approaches.‘

Author: Marius Dorobantu