- ISBN: 979-8265238634
- Enjoyment: 3/10
- Recommendation: 3/10
Aaron Youngren’s recent book Wild and Glorious is ostensibly a practical primer on how Christians should interact with AI. Knowing the author personally, I was hoping that it would be a breath of fresh air and sensibility amidst the generally geriatric status of Christian takes on AI. I ended up incredibly frustrated.
In part this was because I appreciate what Youngren is trying to do. He is trying to think about technology through Christian anthropology, and uproot common misconceptions and mind-viruses that plague our day. I would like nothing more than a human future. So, while reading through the first part of the book, though I immediately knew I would disagree with Youngren, I was interested to see how he would develop his ideas. There came a section in the chapter “The Impossible Singularity” in which he listed several common concerns with AI and his responses to them. These were things like “AI is going to automate all jobs and make humans irrelevant” or “AI is going to kill or enslave us all.” Light stuff. He wrote a bit about each of them, more or less saying none of this is possible, and you only think it is because you’ve been taught to. Then the book moved on. Hm.
At this point I was alert. These are bold claims, and he has only outlined them. Since this was pretty close to the start of the book, I read on with the expectation that he would elaborate on these claims and provide further reasoning. He did not. That was the end of the discussion. Both AI doom and utopia are impossible because the Bible says so. This is a move I like to call the regulative principle of ontology. Youngren plants the seeds for this within the first several pages of the book:
If, searching the Scriptures, we find no place for man-made intelligences in God’s plan for the universe, then any talk of further advances in the technology, any attempts to recommend its use, any discussion of this or that promise of freedom from toil and loneliness and death is irrelevant. […]
If, on the other hand, we find that God’s plan might include man-made intelligences, we have a lot of work and conflict ahead. You can insist that men shouldn’t build something all you want. If, searching the scriptures […] we find that our choice is not between building something and building nothing, but rather between building towers and building temples, then we had better start thinking about what bricks are for. (4–5)
The standard regulative principle of worship is an iff statement: something should be part of worship if and only if it is explicitly mentioned as part of new covenant worship in the Scriptures—only those things and all those things. Similarly, Youngren here is claiming that we should only reason about AI insofar as the Bible speaks to it.
Youngren says the idea that AI will one day transcend us is a false narrative encouraged and made popular by public schools and universities. I don’t think he’s entirely wrong about this. We have been told time and time again that the imminent takeover of AI is inevitable, perhaps even desirable. Youngren is right to point out that it’s a bit fishy that the technocratic elite seem to be that narrative’s biggest advocates. But the best lies are based in truth. You cannot dismiss concerns about AI just because the people shilling it are slimy.
These promises of a workless painless future rest on a gross and hell-bound miscalculation: that we can make something that transcends us. We can’t. It is not possible. It is simply delusional pride to believe that we are capable of creating something as magnificent and glorious as ourselves. It is a farcical impossibility. (22)
Youngren says one reason this is not possible is because God made us in His image, the imago dei, and only He can do that (23). It is not possible for us to create something else in the image of God. But this is irrelevant. AI need not be ensouled to replace us. It’s a common concern among AI folks that we might end up in a world where AIs and emulated minds do everything and yet have no consciousness, let alone a soul.
Youngren continues upon this dogmatic theme later, in “Teeming with Intelligence,” which was for me the most frustrating chapter of the book. He opens by saying that even if you decide you want to get away from AI, you can’t. You might purchase a plot of land in the country to get away from it all, but even then, reclining in a field, you will be surrounded by multitudes of non-human intelligences. So far I am in. I think he’s about to paint some sort of machines of loving grace scenario, or say that even if you try to ignore AI, someone else will build it and you will be subject to the world that they create. Instead, he makes an analogy between AI and animals.
Whatever shape our arguments for or against non-human intelligences might take, any arguments about whether they should exist is off the table. They already exist. The Lord made them. And the Lord called them good. So we can argue about the morality of creating intelligences of our own (though I certainly can’t see any prohibition in Scripture for doing so). We can argue about whether we as a race have the requisite wisdom and maturity to create such intelligences. We can argue about whether such intelligences might be useful, or good, or beautiful. But we cannot argue about whether they should exist. Because they already do. (44)
But of course, “non-human intelligences” is a category founded on negation. Animals and machine intelligences are only similar in that they are not human. But a binary state of having intelligence or not is not the salient differentiator. What is important is the level of intelligence and scale.
Do you know what else are non-human intelligences? Demons. I don’t think AI is inherently demonic, but this shows how the analogy breaks down.
I understand why Youngren grouped these things the way that he did. He’s reasoning from the imago dei point he made earlier. Animals and machines are both intelligent, but do not have the imago dei. This is about where the relevant similarities end.
I do agree with one thing: machine intelligences are wild. We do not know how they work and likely never will. They are less controllable than ordinary software. And in this sense, sure, they are like animals. But I doubt Youngren would like to deal with a troop of 180 IQ bears during his picnic in the park.
This is part of a larger problem of mistaking pastoral advice for claims about base reality. For example, later on he says “[Whatever] benefits man-made intelligences can bring us, an end to labor is not among them” (78). And in another place, he claims that God’s covenant with Noah precludes any future catastrophe:
Even if we granted divine powers of creation to humanity, a nearly endless number of additional objections to this silly narrative [that AI could kill us all] would immediately arise. For starters, God Himself promised that a global, mankind-eradicating catastrophe will never again occur. (30)
But while we might take comfort in God’s promise to Noah, it is not a categorical promise about all possible dangers to mankind such that we can say, “no, that can’t happen—it’s impossible.” Such a stance is self-defeating anyway. Rarely does avoiding catastrophe look like divine intervention. It looks like people getting together to solve a difficult problem. The ozone layer/CFC problem is a prime example. There really was a hole in the ozone layer. Things really could have gotten disastrous. What prevented such catastrophe was everybody deciding to stop using CFCs and work to reverse the damage already done.
These early errors lead to some unfortunate conclusions. Later, when talking about how to deal with potential harms from AI, he says “Intelligences are wild. So is the ox. Yet, the command in Scripture is not to kill all oxen. The command is to use the ox, and to know what to do when he acts unpredictably” (81). But the kinds of dangers here are massively different, and we have no reason to expect we could contain them or perform any meaningful triage. Just look at the Hugging Face incident. This is only going to get worse.
Wisdom would dictate not to allow this eventuality in the first place. But, quoth Youngren (and many on the right), that would stifle development. Instead, he proposes compensatory laws (82). But this is short-sighted. It is only relevant in a very thin slice of history. You cannot compensate for extinction. Youngren, having dogmatically closed that off as an option, does not deal with that at all. This is why this is so dangerous. By ignoring potential risks (despite all the theoretical and empirical evidence for them), you render yourself unable to prevent them.
This is most apparent in the chapter “Skilled in His Labor.” Youngren advocates for cultivating personal excellence and skill in areas where one uses AI. To justify this, he imagines what it would look like to increasingly hand over our workload to AI:
My guess is (market competition being the mother of invention), it would be a short time until this logic would get out of control. Someone would ask the intelligences what microchips we should make for them next. And perhaps for a master plan to optimize domestic power use for a bit more compute. And (while we’re at it) what might be a good surveillance strategy to make sure we’re doing our part? (91)
His conclusion, after this, is that “[dependence] is a quick path toward slavery” (90). This seems to be the default path. In fact, it is already happening.
Youngren comes so close to predicting what will happen, but then seems to think that those who deliberately refuse to become dependent on AI, and instead are masters themselves, will dominate. At least so far, that is not the case. It is an adapt-or-die world. Adapting, in this case, means having an LLM write all of your code. I do think that, ceteris paribus, it is good to demand skill from ourselves. However, this is not the right lesson to take from increasing dependency on AI. The lesson to take is that we will inevitably turn our agency over to machines if we build them, and so we should probably stop doing that. This is a probability problem, not a virtue one.
There is much I agreed with in the book—for example, we ought to fight the Landian Neo-China beast instead of fear it (32). But these things were overshadowed and made impotent by his categorical rejection of serious risks (or transformative goods) from AI. This is a shame, and frustrating since I regard the author highly in his other pursuits. Sadly I suspect it will only deepen the sentiment that Christians are helpless when it comes to reasoning about technology. If we ignore the potential power of AI, we make ourselves unwitting fools to those that wield it.
