If you know me, have followed this blog, or even looked up at the name of this website, it might not surprise you to learn that I really really did not like the new documentary by Chris Paine, Do You Trust This Computer. The title to this post calls it propaganda, and that is not an opinion that I give lightly. There are some seriously talented AI researchers featured in this film (although as I will talk about, with a few notable exceptions, they make much more nuanced claims than the overall tone of the documentary). So while I admit that the fact that Elon Musk gave the funding for it to be shown free last weekend (thanks?) did raise my hackles, I did try to engage it as a serious argument. Unfortunately, as a whole, the film does not make a good-faith attempt to a coherent argument, but rather is only interested in making you as afraid of AI as possible. This a real shame since there are smart people talking about real issues, but these are not really given any focus or clarification and these bits are outnumbered by the pop-culture evil robots, scary montages, and shockingly stupid statements made by people who do not know what they’re talking about.
The documentary is only about an hour and 10 minutes long, but by my count talks about at least nine separate topics which require more than the average here of less than eight minutes of screen time. So in this first post, I’m going to give my overall thoughts on the movie, defend why I called it propaganda, and tackle some of what I think are its worst sins. In subsequent posts, I will go through many of the more specific topics and break down where I disagree (or agree!) with what’s being said and talk about where I think the interesting discussion on some of these topics lies.
First, I’m going to begin with what seems like more of a technical filmmaking critique, but I think is actually pretty important reason why I think this movie is an act of fear-mongering rather than a serious documentary. The pace and editing of this movie are distractingly fast and confusing. In the about 72 minutes of screen-time, the movie goes through user data and privacy, AI learning from data, AI’s effect on employment, AI human interaction, autonomous weapons system, superintelligence, consciousness, AI omniscience, how AI affects politics, and human-AI merging. Any one of those is probably a good subject for an hour long documentary. I have personally had hours-long conversations about just one of these. But rather than really dive into any of these topics, the movie just sort of schizophrenically moves between all these topics, fast-cutting between interviews and inter-cutting with scary imagery or scenes from classic Sci-Fi. It’s really amazing how hard this movie is to follow. In fact, I had to really struggle to come up with that list of topics because the movie doesn’t have any kind of structure you can follow. There are no clear lines between sections, each just leads into the other without any real coherence.
The result of this editing is that you really don’t get any really thoughtful discussion on topics. People talk some words at you, you watch some movie about an evil AI and we’re on to the next topic. There’s no time to think, and that’s the idea. This movie doesn’t want you to think critically about any of these ideas. All it wants you to take away from this is “AI SCARY!” Don’t believe me?
Why this movie is propaganda
Like I said, calling a movie propaganda is a pretty bold claim, and one that I’m not making lightly. But when you look at what the movie does, it’s hard for me to reach any other conclusion.
Also see how I’m making subheadings? That’s called organizing content. Little tip for ya there Chris Paine.
Use of visuals
First, the documentary makes use of incredibly heavy-handed use of montage and visuals to scare the viewer. One of the first things that happens in this movie is an unseen macbook user clicks “Yes” on the prompt “Do you trust this computer”
Literally the next thing that happens is we see the scene from Terminator 2 where a robot crushes a human skull
And that basically sets the tone for the kind of blunt-force hackery we’re dealing with here. There are countless examples of over-the-top visuals being used to try to scare the audience. When they talk about autonomous weapons systems, we get this image:
Yes, that’s some hapless dude presumably about to get get drone-striked by a rogue AI.
But the standout for me is a montage near the end. It’s pretty much just the brainwashing video from A Clockwork Orange. At the end of the documentary, we see a really rapidly edited video that includes a robot, robot bomb, phone, satellite, iphone, rave, another robot, helicopter, guy being carried away from bombing, man in a hazmat suit, flower, geisha, skiing, rollercoster, crowd, robots, drone strike, lava, lightning, tornado, lava. Then it gets all weird and pixelated and puts more weird images in front of us with the eerie pixelization including this one:
Yep, that is a human in what seems to be a ruined civilization with a giant human skull in the background. There really is no subtlety anywhere in this documentary.
Oh, also there’s literally a Hitler and Stalin image in this documentary. But we’ll get to that, I promise.
Hollywood scary AI
Like I said, the documentary plays a lot of scenes from classic to modern Sci-Fi. Terminator, Ex Machina, Robocop, 2001: A Space Odyssey and The Matrix are all featured.
Near the beginning of the movie, Jonathan Nolan is seemingly justifying why they’re doing this.
“I think we completely [expletive] this up. I think Hollywood has managed to inoculate the general public against this question. The idea that machines will take over the world. We’ve cried wolf enough times, the public has stopped paying attention because it feels like science fiction. Even sitting here talking about it right now feels, it feels a little bit silly. Oh this is an artifact of some cheese ball movie. But it’s not. The general public is about to get blindsided by this.”
Maybe I’m reading too much into this quote, but it seems like they are trying to say that comparing to Sci-Fi isn’t a wrong comparison. But even with this justification, cutting in science fiction scenes where robots literally destroy humanity in a supposedly serious documentary about AI seems pretty heavy-handed and dishonest.
Unsupported and wacky claims
The documentary makes really bold and overwrought claims about AI that aren’t really supported. There’s a lot to go through, but here are some of the most egregious for me, in no particular order.
At the beginning of the movie, we get a lot of the punchiest quotes:
“What we’re on the brink of is a world of increasingly intense, sophisticated Artificial Intelligence”
“The robots are coming and they will destroy our livelihood”
“We have a networked intelligence that watches us, knows everything about us, and begins to try to change us.”
“Mass unemployment, greater inequality, massive social unrest”
“Regardless of whether to be afraid or unafraid, the change is coming, and nobody can stop it”
“The technology leaps are going to yield incredible miracles, and incredible horrors”.
Annoyingly, none of these quotes are directly attributed, so I really have no idea who said most of these. Some of these are more or less right, but a lot are just scary-sounding and aren’t really given support by the actual documentary. Honestly, just throwing out so many scary quotes at the beginning with no attribution at the beginning is bad form. Let’s move on to some of the more interesting claims made.
In the part of the documentary talking about super-intelligence, they talk about GoogleDeepmind’s recent work in reinforcement learning. They show the Deepmind AI that can play Atari games, Go, and mention that an AI recently beat top humans at poker. They then make what I think is what I think is a pretty big leap that therefore, AI can learn basically anything given enough time. I want to devote a longer post to this topic, but my basic argument is that while we can teach an AI to learn very specific tasks like games, these systems are really like Clever Hans, and what they learn from one task does not help them with other tasks or be sort of generally intelligent.
Also via Musk:
“[DeepMind] is absolutely focused on creating digital superintelligence, an AI that is vastly smarter than any human on Earth and ultimately smarter than all humans on Earth combined.”
I think this is also a bold claim that’s not really supported by the evidence. For the record, DeepMind’s mission statement reads in part:
“We’re on a scientific mission to push the boundaries of AI, developing programs that can learn to solve any complex problem without needing to be taught how.”
Musk also makes the the claim that DeepMind’s AI is connected to their main server, so if it went rogue, it could have all of the data Google has (which is pretty much all the data). I don’t actually know the specifics of how Google organizes its data, so I won’t call complete BS (although I don’t think Musk does either). But I will say, if Google is running any program (“intelligent” or no) with sudo access to their entire network, maybe humanity deserves to be destroyed.
“We’ve opened Pandora’s box. We’ve released forces we cannot control and cannot stop. We’re in the midst of essentially creating a new life form on Earth.” – Jonathan Nolan, Writer/Director
This has literally become a Greek tragedy folks. I guess we’re just lucky he didn’t reach for Euripides here.
And I swear, I didn’t specifically intend to pick on him, but the the most ridiculous thing that anyone says in this documentary is what Elon Musk says right before the end.
“At this point if there’s an evil dictator, that human is going to die. But for an AI, there would be no death. It would live forever, and then you’d have an immortal dictator from which we can never escape.”
As promised, after this quote is when we see Hitler. So basically, I would only be slightly exaggerating if I said that this movie says that AI is literally Hitler. Actually, I guess this movie is saying AI is literally worse than Hitler. Put that on the box!
Why is Jonathan Nolan everywhere in this movie
Oh hey, did you notice there have been a lot of quotes from Jonathan Nolan. He’s in this movie a lot. Why is that exactly? He is not a computer scientist. He hasn’t written any academic work on any subject. He’s not a journalist or Tech CEO. He’s not even the best Nolan (sorry, cheap shot).
His quotes in this movie usually serve one of three purposes:
- At the very beginning of the documentary, as we discussed, he talks a little about Hollywood in AI
- To transition into a new topic, or provide some kind of transition between interview subjects
- To say something really outrageous that they couldn’t get someone who actually knows AI to say
The first thing is totally fair, although I’m not sure why they couldn’t get someone who was actually credited for writing a Terminator movie. The second one is also perfectly valid. Except for one notable exception, there is no voice given to the director, which is a little unusual for a documentary. So, lacking a narrator, you need someone who can provide these transitions. However, I think the third thing is his real purpose here. By having someone who you can (presumably) get to say all kinds of ridiculous things about AI, since he doesn’t have a reputation in the AI community to protect, the documentary is able to layer in his comments with those of more serious interview subjects. Because of the fast editing and framing of this movie, you really have to pay attention to know who is saying what. The effect of this is that his claims are given equal narrative weight to more serious voices. This, I think, is a really dishonest tactic on the filmmaker’s part.
So what are we all doing here?
Okay, so I said the movie was propaganda. I think I laid a pretty good case that the movie uses a lot of manipulative movie-making techniques and makes a lot of bold, unsupported claims. But what was this all in services of?
I said before that there was one exception to the rule that there was no author voice in the movie. Here is that one exception.
And now everything falls into place. This movie was made to scare us (in a very general way) about AI. If the movie actually wanted to talk about regulation, why was this card at the very end of the movie? Could it be that the director has absolutely no idea what kinds of regulations make sense for the (approximately) nine topics he brings up in the movie? Or that it’s much easier to make a scary movie than to actually make some kind of specific argument?
So those were my overall thoughts. I thought it made for a pretty good Sci-Fi thriller, but if you were expecting a documentary that actually did some kind of meaningful dive into the issues we face as AI becomes better and a bigger part of our lives, you’ll be disappointed. It’s just fear-mongering.
Like I said, there are a lot of important topics that are nominally covered that I think are really important, and I don’t think you can brush these off. I also realize that I don’t go through a ton of specifics on some of the claims made here. So to that end, I will be making a series of posts that address some of these, so look out for those in the coming weeks.
Did you disagree with anything I’ve said? Please leave a comment. I’ll try to address them when I do my deeper dives into some of these topics.