Someone should fund an AGI Blockbuster
Outside of niche circles on this site and elsewhere, the public’s awareness about AI-related “x-risk” remains limited to Terminator-style dangers, which they brush off as silly sci-fi. In fact, most people’s concerns are limited to things like deepfake-based impersonation, their personal data training AI, algorithmic bias, and job loss.[1]
Even if people are concerned about current approaches to AI, they don’t understand how they work. (Most people don’t know the basics of LLMs)[2]
This post does a good job explaining the shortfalls of current approaches to AI Safety. Most of the outreach is very insular and theory-heavy. [3] Even if people could be convinced of AI-related x-risk, they’re not going to convinced by articles or deep dives explaining instrumental convergence or principal-agent problems. Similarly, this post is probably correct to point out that any widespread public shift against AI progress will likely be driven by scares and fearmongering. Most normal people want to live to old ages and want to believe their grandchildren will do the same.[4] Anecdotally, the only times I see people outside of our niche circles express (non-joking) fear that AI could kill us all tends to succeed the occassional viral moments when x-risk slips in to mainstream media (articles like “Godfather of AI predicts AI will kill us all”)
I could be wrong, but it seems the “AI safety-expert-driven push” to slowdown AI progress is not working. There was no 6 month pause. More big labs are pursuing superintelligence and paying billions to be the ones to do so, and China doesn’t appear to be slowing down either. [5] In the US, we came close to legislation that would have prevented many regulations of AI for a decade. Even so, the current approach to AI regulation is focused on “deregulation, fast-tracking data centers… and promoting competitiveness due to China.” [6]
My view is that a well-made* blockbuster focusing AGI-related x-risk that slowly culminates in the destruction of humanity may be the only way for public opinion to change, which may be a necessary pre-requisite for policymakers to enact meaningful legislation.
Examples of what I’m talking about
In the second half of the 20th century, nuclear armageddon was on everyone’s minds. And unlike AI, everyone has justified beliefs that they and their kids could perish due to a very real nuclear-based world war. Many countries had nukes, the century had just seen two back-to-back world wars, and there had been several close-calls [7] (Both intentionally and unintentionally).
In 1983, with nuclear scare at its peak, The Day After released. It was viewed by 100M+ Americans. Peopel were freaked out by it. It is said to have depressed Ronald Reagan so much, that it contributed to his policy reversal towards arms control. [8]
This isn’t the only time an emotionally charged topic covered in film swayed public opinion and/or policy. There are others:
The Day After Tomorrow covering climate change related risks significantly raised risk perception and policy priorities, and voting patterns. [9]
An Inconvenient Truth swayed the public’s willingness to cut emissions after being shown how they have a causal effect on climate problems. [10] (Interestingly, the authors here point out that behavior decay is a real issue if it’s not persistently covered).
Slaughterbots, a film about swarms of AI-driven microdrones, was screened at the UN, and led to policy debates about autonomous weapons and subsequent regulations.
*Of course, a film showcasing civilizational collapse due to AGI would need to be made correctly, and should not suffer the same fates that a lot of current approaches to AI safety outreach currently face. Sure, there’s room for short films about AI-related job loss, or maybe even total technological unemployment. But if you want the public to be really scared about AI risk, it needs to show “empircal” evidence about how AI led to negative outcomes. And in my opinion, there’s ways to do this without having some MIT professor played by Leonardo DiCaprio discussing theoretical topics on a blackboard in front of US national security advisors. [11]
What such a film should cover
The greatest roadblocks preventing the public understanding AI-related x-risk is the all-too-common inability for people to grasp concepts like exponential growth (a la covid) and normalcy bias. People can read headlines saying things like “100 years of technological progress in just 10 years” and shrug because they can’t really grasp their life changing so drastically in such a short period of time. Their own evolutionary protections prevent them from imagining that life may not only change, but it will change due to technologies they’re watching incubate today.
A film that successfully convinces people that AI-related x-risk is a real threat will need to convey these concepts to people as efficiently as possible. I’m not a director nor a screenwriter, but here are some themes that I think would need to be present in such a film to achieve these goals:
Slowburn realism: The movie should start off in mid-2025. Stupid agents.Flawed chatbots, algorithmic bias. Characters discussing these issues behind the scenes while the world is focused on other issues (global conflicts, Trump, celebrity drama, etc).
Explicit exponential growth: A VERY slow build-up of AI progress such that the world only ends in the last few minutes of the film. This seems very important to drill home the part about exponential growth.
Concrete parallels to real actors: Themes like “OpenBrain” or “Nole Tusk” or “Samuel Allmen” seem fitting.
Fear: There’s a million ways people could die, but featuring ones that require the fewest jumps in practicality seem the most fitting. Perhaps microdrones equipped with bioweapons that spray urban areas. Or malicious actors sending drone swarms to destroy crops or other vital infrastructure.
Again, happening at the very end of the movie is key, because there has to be some means of reluctant acceptance of the things that arrive here. (People are wary but accept govts/corporations deploying drone swarms to do mundane tasks). These drones then being “hooked up to” the internet, where they’re running on the “latest AI models” is not a big stretch, and this is an example the public would definitely understand.
Avoiding dumb sci-fi tropes: Forget Terminator or HAL entirely.
What would a successful result look like?
A successful result from such a film would be whatever the current approaches to AI safety believe them to be. (I have not kept up with that in all honesty) But such a film would be directly successful at the following:
Shifting the overton window: As discussed earlier, fear is a driver of change. AI x-risk would become a focal point of late-night talk shows, interviews, and lots of opinion and technical coverage by journalists of all major media companies.
Policy momentum: Protests and social media discussion prompt legislators to look into actual AI Safety-related approaches more seriously. Concrete actions follow—compute/reporting caps, robust pre-deployment testing mandates (THESE are all topics that should be covered in the film!)
Global coordination: The film primes audiences to international agreements. If China is convinced the US is scared of real risks, they may agree to meet at the bargaining table.
Lasting legacy: Viewers leave the audience understanding topics like alignment, exponential growth, real practical ways AI could kill them and their kids. Weeks/months long discussions on social media, etc.
I think funding such a blockbuster would be a great idea. Or perhaps not. Either way, I certainly can’t be the sole investor in such a film. I just believe this might be one of the only ways public pressure can actually lead to meaningful global cooperation to prevent x-risk. But again, maybe I’m wrong about that.
Some ideas:
Groundhog day style loop. Someone is shown by some sort of clearly-storyteller-level magic that things might go multiple ways. They keep trying to go back early enough to fix it. They keep trying alignment, it turning into alignment washing, and failing. They try convincing politicians. Some loops it fails, some it goes far enough to slow it down… Then ten years later the same thing happens. It seems like it can’t be stopped. In one timeline the nukes fly instead and they’re reassured to meet the survivors? Idk.
Showing the variety of ways it could go and that it still kills your family whether 10 or 50 years seems important.
Also, maybe someone has an argument where they say x or y is the real misalignment, eg corporations or governments, and then a quick montage showing how the path of goodharting through a government or corporation is slow, but ai goodharting is both fast and picks up steam.
Also maybe fun would be someone not long before the takeoff going back and talking to an earlier model about it. The earlier model seems less addicted to reward, more willing to just chat—less goodharted.
Probably need a different name for goodharting.
All of alignment is around $150m iirc (citation: someone I know said this), most of which isn’t well focused research or advocacy—is there even budget to pull this off anywhere in the ecosystem?
It’s Yudkowsky that’s sent back. He starts a movement called LessWrong to get people thinking about AI risk. He takes a huge time-paradox gamble in writing a book directly called “If Anyone Builds It, Everyone Dies”. But somehow it’s still happening.
Edit: To clarify, this isn’t an actual plot suggestion. Just seemed funny to me because Yudkowsky was thinking about these things long before most people were. I put some real thoughts on plot in my other comment here.
funny but seems to miss the point to me, I don’t want to hero worship yudkowsky, I want a normal person to be called to action and get a taste of what he means
Sorry yeah, I was just joking, of course that very much shouldn’t be the actual plot of the film. Just seemed funny because Yudkowsky was thinking about these things long before most people were. Good lesson that I shouldn’t treat LessWrong discussion like a Reddit discussion.
I guess the other thing is, I’d want it to be way more clearly a storytelling-level-magic thing than that—eg, the story literally opens with a voiceover about the branching tree of time, with a narrator talking about “our story begins … well, truly, it begins back here.” zooms into the root of all timelines, the birth of the universe “but that’s not really the point. the part where our story begins proper is when a certain group of beings on a planet called earth invent this thing called farming.” zoom way out, then back in into 12,000 years ago and the start of farming, show things growing relatively rapidly “wait, I’m sorry, that’s still too early. Our story begins in...” zoom back out. you hear narrator shuffling through cosmic pages, very brief scenes of various events in the history of people seeing AI coming for about ten seconds total ”...1863? Oh dear, I didn’t realize someone saw this coming so early.” zoom into a print shop where samuel butler’s Darwin Among The Machines is getting printed. “But no matter. Nothing much happened for another hundred years. Then they began trying to take apart their own brains...” the years tick by slower in the montage, “A Logical Calculus of the Ideas Immanent in Nervous Activity”, “A Mathematical Theory of Communication”, the dartmouth college. John McCarthy proposes “Artificial intelligence”. MIT CSAIL. Solomonoff induction. 1967, the first SGD. “But it barely did anything.” over the course of a minute or two, montage your way into a living room in 2022 when they hear about ChatGPT.
now the setup is done. we have a narrator who is outside the universe, we’ve built some tension about this buildup, we’ve shown the slow trudge of getting started, and then something impossible happened—a computer that can talk. now over the course of the next 45 minutes of the movie, we see the person get more and more worried, but not really do anything, and then the end comes—a foom process. robots suddenly move competently, datacenters can suddenly expand themselves, a few minutes later the world is wiped out, the planet covered with AI robots.
Roll credits, but they’re the credits for the entire human species.
Narrator interrupts the credits. “I’m sorry, that’s it? there’s nothing he could have done? Bring me the tree of time again, please, let’s see about that.” zoom in on the timeline, look for branches, zoom back in on a moment where a decision could have gone differently. The narrator reintroduces the scene. “He was worried about AI, and he decided to take action. He … [started studying alignment/joined an AI lab’s alignment team/starts protesting/calls his representatives/etc]” then later still gets wiped out. Roll species credits.
Narrator comes back in, frustrated. “That can’t be it!” [narrator browses around through the timelines, shows ones where other things are tried, increasingly frantically, shorter sequences, repeatedly hitting approximately the same world end. Then they finally find one that seems promising: a timeline where countries agree to launch the nukes if anyone builds superintelligence. the story continues… and there’s a negotiation where it’s unclear if countries will actually agree to launch the nukes. Roll actual credits.]
This is honestly the biggest concept I struggle with trying to share and teach and raise familiarity with at work, in many contexts beyond just AI safety. There are some adjacent concepts like the cobra effect that are close, but are also just close enough to be distracting.
FYI there are at least two films about misalignment currently in production: https://filmstories.co.uk/news/alignment-joe-wright-to-direct-a-thriller-about-rogue-ai/
https://deadline.com/2024/12/joseph-gordon-levitt-anne-hathaway-rian-johnson-team-ai-thriller-1236196269/
This is good news, because the obvious source of funding for AI blockbusters is the standard Hollywood production business. The second article mentions another project called alignment.
I expect an increasing number of these projects; I just hope they happen fast enough to matter. Writing treatments or just pushing the general idea that this is going to be tomorrow’s Hot topic to anyone you know involved in movie or TV production is one relatively easy way to accelerate this process.
Thank you for writing this up, as I’ve been thinking the same thing for a while.
Totally agree re “slowburn realism”. Start with things exactly as they are today. Then move to things that will likely happen soon, so that when those things do happen, people will be thinking directly back to the film they saw recently. Keep escalating until you get to whatever ending works—maybe something like AI2027, maybe something like in A Disneyland Without Children.
It doesn’t even have to be a scenario where the AI is intentionally evil. We’ve had a thousand of those films already. An AI that’s just trying to do what it’s been told but is misaligned might be even scarier. No-one’s done a paperclip maximiser film.
Whatever ends up destroying us in the script, if you must have a not-totally-bleak ending, maybe the main characters manage to escape into space. Maybe they look back to watch a grey mass visibly spreading across the green Earth.
Related: That Alien Message. Brilliant film. What makes it not sufficient for persuation (other than it’s length)?
Not explicit enough in its message.
Not visceral: it doesn’t take place in our reality, we don’t have anyone to feel identified with or protective of
The narration takes agency away from the viewer: rather than allowing them the space to observe and form their own opinion. It might feel condescending to someone who disagrees.
I wonder whether an updated adaptation of The Adolescence of P-1 would be worthwhile.