Chris Watts Article

P(Doom), AI, and the Christian

Introduction

Artificial Intelligence (AI) has subsumed the American economy; the majority of stock market gains and capital expenditure come from this one sector, driven mostly by seven major AI companies. It seems like every week there is a new major advancement in AI capabilities or some major AI story:

It’s been a while since I wrote anything substantial on AI, and amid all these (very) rapidly developing situations, I thought it might be good return to the topic, and to focus in particular on one aspect of the AI discourse: p(doom). If you spend any amount of time in AI discourse spaces you will have come across this term, but I think it has broken containment several times in viral stories or articles, and deserves unpacking.

What is p(doom) and Where Does it Come From?

If you are familiar with the AI discourse, you will have come across this term, which stands for probability of doom. It has become the shorthand for how people categorize or evaluate the risk of AI, as in “I think there is a 10% probability of doom in the next three years” (it is usually a percentage accompanied by a measure of time). Of course, “doom” is not a very precise word. P(doom) is a useful and memorable shorthand, but it isn’t very specific. In my experience, most people who use the term define “doom” as humanity being exterminated, or at least as modern civilization collapsing and humans returning to some version of technologically deprived hunter-gatherer existence. Your personal p(doom) – how likely you think doom is to occur over some period of time – is influenced greatly by what kind of doom is being discussed, and over what length of time.

This is, in a way that bizarrely mirrors religion, a kind of technological eschatology, or study of the end times. Eschatology will of course be a familiar idea to the Christian; from the very beginning of proclamation of the Gospel, Jesus and his followers have taught about and thought about the end. The last part of the Bible is predominantly concerned with eschatology, but eschatological thought is scattered throughout the New Testament and indeed is present throughout the Bible as a whole.

For the Christian, then, p(doom) functions differently than for the secular technologist in one key respect. Rather than an estimated percentage over a given time-frame, Christian p(doom) is 100% over an entirely uncertain time-frame. Doom will occur, but not in any sort of predictable or measurable time. The likelihood of doom tomorrow is exactly the same as the likelihood of doom 100 or 1,000 years from now; doom is completely certain in the aggregate, but completely uncertain in the specific. Because of this, AI p(doom) discourse is both strange and familiar. Christians of various flavors have been (wrongly) predicting the end for as long as Christianity has been around, so the AI doom discourse may not seem that strange to you (as in, “what’s the big deal?”), depending on what kind of Christian community you grew up in or currently inhabit.

But the technological version of p(doom) is dramatically different from Christian eschatology, for several reasons. The first difference is in who is driving the discourse; the people currently talking about the end times are distinctly different from normal eschatological types in that they are typically non-religious (at the very least non-Christian) humanists who have, up until the last four years, widely and loudly proclaimed the blessings and goodness of technology. Many of the same spaces and communities are simultaneously predicting that we can solve human aging in our lifetimes, and that AI will kill us all. The second difference, as we have already stated, is that these claims involve supposedly definitive time-frames; in large sections of the AI research community, the question is not if doom will come, but how soon.

The third, and most bizarre aspect of the doom discourse is that the people who are most concerned about doom are the very people building the doomsday machine. This is why the most recent AI researcher resignations have made such waves. The people building AI are the very ones who are most convinced that AI will probably kill us all.

This might come across as extremely irrational or unusual behavior. If they are so convinced the thing they are building will kill us all, then why are they building it? The AI sector is trapped in a vicious doom loop (pun intended) that has resulted from several convergent factors:

  • There’s a version of the sunk-cost fallacy at play here; we have already poured so many resources into AI that it would be foolish to stop now.
  • Because of the vast amounts of money that have been spent on AI, there is immense pressure to keep going until the AI products are profitable, so that all the people who have invested so much get some sort of return on investment.
  • This is also being driven largely by fear. Yes, these machines might be terrifyingly powerful, so what happens if someone else gets there first? What if China (or a rival American company) makes the world-ending technology before we do?!

You might wonder if the doom discourse is insincere nonsense intended to hype up a product (“our product is so powerful it might kill everyone, think of how it could help your business!”). But I don’t get that sense in reading statements put out by the labs or by various critics or commentators. I do think that many of the people at these labs do genuinely thing that AI could kill us all. Thus, you might think if these people think AI will kill us all, why are they still building it? The answer is pretty boring: they are doing it for the same reasons people have always done things.

For the most part, human behavior is driven by incentives; people act in the ways that they think will most give them what they want. But human incentive structures are not entirely logical; people often (or maybe even mostly) want multiple contradictory things. Despite our best efforts, no one behaves entirely rationally, even the people most devoted to rationalism. We are all bundles of competing desires that pull us in different directions. It should not be surprising to the Christian that AI companies are continuing to build the doomsday machine; one of the core tenants of Christian teaching is that people behave in inconsistent, self-defeating ways all the time because everyone (including Christians) is subject to the fallen nature of the flesh.

Money (greed), fear, and pride have always led people to behave inconsistently or irrationally. These are all compounded when you are considering not just one person but many. There isn’t just one person in charge of the AI build-out. It is driven by a multi-faceted conglomeration of researchers, investors, companies, and politicians. In the aggregate, people can be even more irrational than individuals.

This, then, is what is behind the p(doom) discourse you may have seen; when you see this terminology online you will hopefully have a better idea of what exactly is being discussed and why. But there are several other layers to this idea that I think are worth exploring.

How Should We Define “Doom”?

Christians can reason ourselves out of the most extreme versions of p(doom): AI simply can’t be an existential threat to humanity, for a variety of reasons. I have previously outlined these in detail, so a brief summary will hopefully suffice.

  • All the Scriptures that talk about The End assume that humans will be around for Jesus’ second coming
  • Things will continue along basically as they have been until Jesus returns (note that this does not mean things will be good when Jesus comes back; war, upheaval, and death are the normal state of affairs for humanity).
  • Thus, if the Bible is true and literal (as I suspect most of my readers believe) and the writers were inspired by God who knows the future, we can safely rule out AI as an existential threat to humanity. In other words, I am very confident AI won’t kill all the humans.

But, as we have said, “doom” is not a very specific word. There are many possible outcomes of AI that might not technically be existentially threatening, while still being very, very bad. The fact that humans will still be around when Jesus returns doesn’t mean that our current civilization will be around. Human history is not one of perpetual forward progress, despite what many political and social commentators might have you believe. Human history is littered with dead and fallen civilizations. Things don’t always get better, and states of affairs in the world don’t last forever.

So, my p(doom) might be 0 for total existential risk, but my p(doom) for “doom=societal upheaval” or civilizational risk is not 0. It’s more like 20% over the next decade. There is an excellent book called Guns, Germs, and Steel that I think provides a pretty good framework for thinking about AI risk, though that book was written long before the current advances in AI and isn’t really about AI at all (it’s about why some societies rose to prominence over other societies throughout history).

Steel (Unequal Technological Distributions)

Most new technologies are mostly amoral, but they are almost never equally distributed, nor used for wholly good or even amoral purposes. Consider: the internet is a great equalizer of information availability, but it is also the technology that underpins the nightmare machines that drive us all crazy (smartphones), and especially in its early days it was very unevenly accessible. At the present moment, the monetary gains to be had from the internet are vastly unequal in their distribution. In my estimation, the most likely “doom” that AI will bring about is this kind base-line technological upheaval.

For example, what if AI breaks public encryption, like WiFi or Bluetooth? What if models become cheap enough to run locally that anyone with even modest access to resources will have the ability to break most consumer-grade digital security? I don’t think it’s unreasonable to speculate that AI might soon (within the decade) have this kind of capability. This is the backdrop of one of the recent AI stories that went viral, concerning Claude Mythos.

The natural response is “well, won’t AI also enable better encryption and security, and mostly leave us in the same situation we are in now?” Maybe? But it only takes offensive capabilities running a little ahead of defense, for a short amount of time, to create absolute chaos. And the defensive capabilities almost certainly won’t be equally distributed in society. Maybe governments and central banks will be able to withstand the AI onslaught, but the average consumer will be totally vulnerable. It is very possible in such a scenario that some governments will be better prepared than others; whole new kinds of warfare and espionage open up to governments with frontier AI models, at the expense of nations who don’t.

This is just one example of the ways in which the unequal distribution of new technologies affects societies and causes upheaval. Maybe AI won’t break the internet, but will further fracture and deteriorate dating and romance, as more people pair up with AI companions instead of with real human peers. My point is that new wide-scale technological adoption typically affects society in uneven and unpredictable ways, and even if AI never becomes “superintelligent” it still might do this to our society.

Guns (Unequal Abilities to Impose Our Will on Others)

Tied to this kind of technological upheaval is the threat of violence. One of the scariest “reasonable” possible outcomes is that AI could enable smaller and smaller actors (groups of terrorists or individuals, instead of governments and large-scale corporations) to engage in biological warfare. It is possible that AI enables the creation of novel biological weapons at cheaper costs than ever before, allowing bad actors access to such weapons at scales we have not previously seen.

AI doesn’t have to get powerful enough to kill us all on its own; to bring about many kinds of “doom” it just has to get powerful enough to enable people to kill each other in new, more powerful ways. Again, I don’t think this will lead to the extinction of humanity, for the reasons outlined above, but new kinds of wars or warfare can still be a kind of widespread doom that leads to the realignment of the current world order.

Germs (Hidden Inequalities that Devastate)

I have previously written about the disempowerment problem in AI research, and this is yet another somewhat likely scenario that could lead to a certain kind of doom. New technologies can lead to political, societal, and cultural upheaval without being existential or even particularly violent. What if AI becomes advanced enough to take over a lot (even most) of knowledge-economy jobs? What do all the people that currently have those jobs do, in that case? Are there enough blue-collar (think electrician or plumber) jobs to go around? What if we solve robotics to the point that those jobs are also cheaper AI to do? Are we, as a society, willing to realign the way we think about the economy and markets?

The problem is not the realignment of the economy, per se; many of the most common jobs today didn’t exist 200 years ago, so we shouldn’t necessarily fear shifts in job markets. Rather, many of the problems AI poses concern speed and rate of change. Economies can shift over generations without a lot of societal upheaval or violence, but if a culture or economy changes very rapidly, within the span of one generation, this often leads to violence, unrest, and war.

Maybe this won’t happen. Maybe we have reached the practical limit of AI effectiveness in the economy. But even AI stagnation poses as serious challenge; if most of the American economy is being propped up by seven AI companies and their insane amounts of data-center spending, what happens when that bubble pops?

Biblical Perspectives

My generation, in this country, has grown up in relative peace and security, historically speaking. This country emerged even from two World Wars in generally the same form as it began them, while large stretches of the world were in many ways remade. There have been border changes, but for the most part the map has stayed remarkably stable for most of our lifetimes, and so it’s tempting to think that humanity has arrived at the end of history; the way things are now is the way they will generally continue forever.

To the Christian, this should be obvious nonsense. Only one Kingdom can and will last forever, and all the nations you can see on a map aren’t that kingdom. Christians must remember that all earthly, human endeavors have an expiration date that only one King knows.

AI has the potential to remake the current world order, either through the distribution of radically transformative technology, or through the enabling of new devastating kinds of warfare, or by rendering our current way of living untenable. But even if AI doesn’t do any of these things, doom is coming for every society that exists, in one way or another. And not just the big Doom at the end of time, but the smaller dooms of civilizational change, upheaval, violence, war, and the redrawing of nations.

Thus, Christians shouldn’t fear AI any more than we fear any other kind of potential future technology. Yahweh is aware of what the frontier labs are building. He is capable of using AI to accomplish his own purposes and enact his own will. God will work though AI for the good of his people, like he does in all things.

But it feels like many Christians are forgetting Jesus’ teaching on priorities:

Do not lay up for yourselves treasures on earth, where moth and rust destroy and where thieves break in and steal, but lay up for yourselves treasures in heaven, where neither moth nor rust destroys and where thieves do not break in and steal. For where your treasure is, there your heart will be also. –Matthew 6:19-21

I think that many Christians are in danger of “laying up treasures on earth” when it comes to the AI discourse. Either people are so attached to the way things currently are (technology, culture, politics, etc.) that they are too afraid of the changes AI might bring, or they are so enamored with AI that they can’t see the potential dangers in the technology. Neither attitude is good! Don’t get so obsessed with the AI discourse that you end up putting your hope in any earthly thing, be it a government, a way of living, a technology, or a national identity.

I don’t know if AI will upend society as we know it, but I am very confident that something will, eventually. Our society will not last forever, because nothing can, except for the Kingdom of Christ. Our ultimate hope is in Jesus and his promises, not in any specific way of living on earth. AI might totally change the way we live, but Christians should understand that even if it does, Jesus is sill in control, and our mission won’t be fundamentally different in the new world order. In fact, we can envision scenarios where the advent of society-altering AI programs makes the Christian mission easier in some ways, not harder. What if AI breaks the internet, and digital life? What if people are forced back into more embodied forms of community, where local congregations are waiting to provide meaning, belonging, and purpose? What if we are forced to build more relationships in person, with our neighbors and coworkers, and thus are given more opportunities to share the life-giving message of Jesus? What if God decides to use the disruptive power of AI to get us all off the doom-rectangles in our pockets?

Is this what’s going to happen? I don’t know, and neither does anyone else. What I am trying to get across is that God can and will use AI for his own purposes, which are often hidden from human understanding. Don’t assume you know how AI will progress, or what changes it will bring about in our society. Continue to trust that God knows what he is doing and will enable his Church to continue to carry out his mission in the world. But, we also can’t ignore AI; we must exercise discernment in discovering how it can be used to proclaim the message of Jesus and grow his Kingdom.

So What?

Many people are already viewing AI and AI risk as sure signs of the Biblical end times, but I would strongly caution against putting much stock in these kinds of claims. AI is not the Antichrist, and OpenAI is not the beast of Revelation (or vice versa). Jesus was clear that his second coming would be totally unpredictable, and that’s why we need to be perpetually ready for his return. Christian p(doom) is 100% over a totally unknowable time-frame, but we shouldn’t get sucked into reactionary or inflammatory ways of equating AI with the kinds of imagery we see in Daniel or Revelation. Don’t be so concerned with the end that you forget to be concerned with how God expects you to live now, in this moment, affecting and reaching the people around you.

To that end, I want to give four pieces of advice for the Christian thinking about AI.

  • Learn what it can do. I think many Christians still have their “heads in the sand” as it were about the capabilities of AI. While it might in theory be a “glorified autocorrect” in practice that isn’t true at all. If you looked into AI several years ago and drew your conclusions about what it is and what it can do in 2022-2024, your worldview is woefully out of date and needs updating. And, practically speaking, if you want to be able to minister to people in your cultural context, you need to know how AI can and might be affecting the people you are trying to reach. If you don’t know where to begin, I’m sure someone in your congregations can help you start out. If there is interest, perhaps I will write more about it on this site.
  • Don’t be like the “simple” in Proverbs. I cannot stress this enough; if you haven’t been keeping up with the current capabilities of AI, you are drastically unprepared for how it has already changed digital life. Even if AI doesn’t break the internet via cybersecurity or hacking, it has already kind of broken the internet by enabling the creation of whole articles, photorealistic images, and video out of very small amounts of base inputs. I strongly encourage people to have a “default skeptic” attitude toward text, photos, and videos on the internet now. If you don’t have good reason to trust the source, almost anything you see online could be an AI creation. When Christians uncritically share fake photos or videos on social media, this is incredibly damaging to our ability to effectively reach the lost. You don’t have to like AI or use it, but you do need to be familiar enough with it that you can reliably spot AI content, or know how to check.
  • Don’t live in fear. Christians have been given a Spirit of power, not fear. Fear clouds our ability to think and act rationally, in control, and with grace and peace. Because of our faith in Jesus, we don’t have to live in fear of AI. When we let our thinking be dominated by fear, we will lose our ability to effectively engage with the world and reach the lost who need the fear-removing message of Jesus. I worry that many Christians refuse to engage with AI at all because they are afraid, and in so doing lose the ability to meaningfully or accurately talk about an important part of our modern culture with the people around them.

The master commended the dishonest manager for his shrewdness. For the sons of this world are more shrewd in dealing with their own generation than the sons of light. –Luke 16:8

Behold, I am sending you out as sheep in the midst of wolves, so be wise as serpents and innocent as doves. –Matthew 10:16

  • Even though we don’t have to live in fear, be aware that change is coming. Remember, even if it’s not because of AI, the way the world is now won’t last forever. But I believe that AI will dramatically alter the ways we relate to each other and the broader world. Christians will be forced to come to grips with the technology, like we were with the printing press, cars, and the internet. This kind of change is inevitable, and we should never be so attached to any worldly way of being that we aren’t willing to figure out how to be a living sacrifice in whatever culture or context comes next. We need to be willing to exercise discernment and figure out how to be pleasing to God in whatever situation we find ourselves in, knowing that God can help us in any circumstance, including any doom that might come from AI.

What then shall we say to these things? If God is for us, who can be against us? He who did not spare his own Son but gave him up for us all, how will he not also with him graciously give us all things? Who shall bring any charge against God’s elect? It is God who justifies. Who is to condemn? Christ Jesus is the one who died—more than that, who was raised—who is at the right hand of God, who indeed is interceding for us. Who shall separate us from the love of Christ? Shall tribulation, or distress, or persecution, or famine, or nakedness, or danger, or sword? As it is written, “For your sake we are being killed all the day long; we are regarded as sheep to be slaughtered.” No, in all these things we are more than conquerors through him who loved us. For I am sure that neither death nor life, nor angels nor rulers, nor things present nor things to come, nor powers, nor height nor depth, nor anything else in all creation, will be able to separate us from the love of God in Christ Jesus our Lord. –Romans 8:31-39