Here's what I've learned from making TikTok videos every day for the last 60 days.
I think posting on TikTok is worthwile whether you want to spread AI safety arguments to an enormous audience, you have something you're passionate about you want to share, or you just want to work on your ability to speak confidently and capture human attention.
There's a list of tips at the end of this post, but first I want to talk about TikTok at a high level. TikTok and LessWrong operate very differently, so even if you are great at thorough, written content, you might find some of the things I learned about short-form video surprising.
I'm not gonna share my TikTok because I'm embarrassed, but here's some info about my account since I started two months ago:
- Highest viewed video: 78.6K views
- Average views over the last 10 videos: 2071 views
- Average views over my first 10 videos: 563 views
- Topics: AI safety, speaking/Toastmasters, interesting facts, AI tips, social anxiety, random fun stuff
My experience creating short-form video was one of regularly being surprised. Videos would do well or flop basically at random. This was a good sign, and part of why I extended my 30 day TikTok challenge to 60 days: making videos every day felt like I was interfacing with reality, and learning every day about human attention and interest from an ~optimal source. In most domains, you would be so lucky to have access to the immediate feedback you have on TikTok. You can be boring for years on LessWrong, and people will politely put in the effort to understand your argument (sometimes). TikTok audiences essentially yell "BORING!" in your face literally 3 syllables into your argument if you're not doing a good enough job capturing their attention.
On TikTok, your followers don't matter. A brand new account can get millions of views on their first video, and an established creator can fail to break 300 (failed videos stop being shared by the algorithm at around 300 "test views".) This means you can learn very quickly how to capture people's attention.
Compared to LessWrong, I'd say the main difference is that people on LessWrong are used to making reasoned arguments, but people on TikTok are usually not reasoning at all. They are turning their brains off when they watch, entering a sort of dissociative trance. Think of it like trying to talk to people who are dreaming, Inception-style. The reasoning faculties simply aren't operating. "High status person said it" or "lots of people are saying it" are evidence, but logic is not. Your video will not modify human nature such that this is not the case. Be prepared for the lowest quality disagreement you've ever received, akin to people talking in their sleep. Youtube comments have famously always been terrible, but TikTok comments make Youtube comments look like the annus mirabilis papers.
I have been watching some Robert Miles on TikTok, and I've been embarrassed on behalf of humanity for how bad the comments on his (thoughtful, intelligent) videos are. Here are some examples of popular comments from his recent videos:
"Oh no I left the door unlocked and look at what the AI did." [Referring to the OpenAI rogue AI model hacking Hugging Face]
Did not happen. Story is faked for hype. [Same incident.]
The chances of an LLM being conscious is absolute 0. By design an algorithm cannot be conscious. No matter how advanced the reinforcements gets its still all simple gates no intelligence at all.
Notice the utter lack of attempt to persuade. These are people who find the existence of their own opinions convincing in itself. The poor quality of comments has actually started to dissuade me from making persuasive content on TikTok. I'm still exploring what kind of content I want to make, but anything political/persuasive seems to invite toxicity and the most frustrating, blind stupidity you've ever seen. If you want to make AI safety content on TikTok, for example, you will need to have a thick skin to this kind of thing.
Making videos, I went through a few phases:
- Learn to stop mumbling. People don't want to watch unexpressive, boring people.
- Improve my lighting/audio. People don't want bad video, and they especially don't want bad audio. I haven't bought a camera or anything, I just got better at using my phone camera.
- Learn to hook attention in the first 2 seconds. I'm hesitant to even say 2 seconds instead of 1. A person might scroll away in a fraction of a single second if the first frame of your video looks like the first frame of a boring video.
- Learn to keep the video engaging throughout, and satisfy the promise made at the beginning of the video. Videos that satisfy people outperform clickbait. (Still working on this.)
- Figure out what kinds of content I enjoy making (Still working on this too.)
Here are some tips I learned along the way:
-
DON'T BE BORING. Cut out s and boring shit. You can lose a viewer in 0.1 seconds of boringness. s or other boring stuff is a strong signal to the viewer that you're an inexperienced creator and the rest of the video will be boring too.
-
Some topics are inherently interesting. Given the same quality of video, a video about Pokemon will do better than a video about Digimon, because people like Pokemon more than they like Digimon. A video about a rogue AI escaping a sandbox will do better than a video about mechanistic interpretability. If you want to teach about mechanistic interpretability, put it inside a larger topic people are more interested in, like feeding a dog a pill wrapped in peanut butter.
-
People usually want to hear a human voice and see a human face. Narration without a face will do worse. No human voice probably flops.
-
Some visuals are inherently attention capturing. People on TikTok use "visual hooks", where they move the camera or their hands in a way at the very beginning of the video to hypnotize you into paying attention for a few syllables longer. If your eye has something else to look at besides the person talking (B-roll, a moving hand, a colorful background), that fills in the gaps of attention where you could start to get bored, and keeps you watching longer.
-
Videos that capture attention at the beginning and then become boring will flop. You have to make the whole video interesting. Completion rate is a big part of the algorithm.
-
TikTok is all about emotion. Your voice, your face, it should all contribute to the emotional message. You are an actor, and literal acting advice is helpful. There is an actor named Sonya (@camerasmartcreator) who gives good TikTok tips from the perspective of an actor. I have also used a book designed for actors to improve my voice ( Freeing the Natural Voice by Kristin Linklater), and have seen substantial improvements in my speaking voice from it. You are producing an entertainment product. Would you watch a TV show where all the actors are quiet, mumble, and speak in a monotone?
-
Sometimes you should just follow your heart and make something you think will flop, just for the fun of it. You might be surprised how often the TikTok audience is willing to join in the fun with you. I did a Slavoj Žižek impression I was certain would flop, and it was one of my most popular videos.
-
I would caution against trying too hard to optimize things like keywords and tricks of the algorithm and stuff that don't matter as much as making great content in the first place. I can tell a lot of people get stuck on "how do I get views" vs. "how do I make content that will naturally get views because it's awesome."
-
If you don't make yourself cringe and embarrass the shit out of yourself, you're not doing it right. At first, I cringed at any attempt I made to be humorous, be overly serious, be loud, or have any facial expressions. These feelings were at odds with reality, and just a normal part of the process. You don't have to get over them all at once and become Mr. Content on day one. Just make gradual progress. If you're not cringing, you're not growing!
-
Make the viewer curious about something that will happen later in the video. The shallowest version of this is the classic, "you won't believe what happened, I'll tell you at the end of the video," but you don't have to be so coarse about it. Just make sure there is some kind of promise of what value the person will get later in the video.
-
Deliver value. That can be entertainment value, that can be a tip they can apply in their own life, whatever, but you have to give them some kind of value. "Here is some information you might prefer to know" is great in neuroatypical communities like LessWrong, but doesn't really fly on TikTok. You're competing with people who promise to help the viewer lose weight, cure their disease, make them laugh, teach them to make money, teach them a mindblowing fact, or show them a scene from a new TV show they should watch. You need to give them value too or you will lose the competition and your video will flop. Would you watch your video if someone else made the same thing? Here's a note I left myself after making a video:
-
" Why joining a club isn't a good way to make friends. Failed. My opinion on loneliness and making friends might be intellectually interesting, but doesn't provide much value for the listener. Delivery was also particularly unemotional and flat. My next video needs to deliver clear value."
-
Opening with a joke is hard if you don't signal somehow there's a punchline coming. The setup to a joke is usually boring and not enough to keep someone interested.
-
If you want to talk about an expert subject, you should have some credibility. "As a small business owner..." Forget what you know about logical fallacies, and appeal the hell out of that authority.
-
Ragebait works. You could use it as peanut butter for your pill. I'm sure you find it as unpleasant as I do, but maybe consider using hints of it in moderation.
-
A video with a surprise in the middle will do well. Or more than one surprise. If I can predict — like an LLM — what the rest of your video will say, why would I need to watch it?
-
A failed video is okay if you enjoyed making it. You have permission to sometimes make something just because you like it.
-
Keep original research to a minimum. You are not required to reinvent the wheel. It's okay to take someone else's concept and explain it in your own way or put your own spin on it. You can't come up with a mindblowing new idea every day.
-
Truth won't stop bullets, nor will it get views (except if the truth is, by happenstance, the kind of thing to go viral. Think Watergate, not Georgism).
-
Content that looks platform-native does better. That means if you have to choose a font for your captions (Use captions! Majority of viewers are muted!), choose the common few that get used on the platform. Don't put weird filters on your voice. If you're talking to the camera, make it look like a TikTok where someone is talking to the camera, rather than content filmed for a different platform.
-
Learning by making 60 videos will take you a lot farther than learning by analyzing, planning, or reading this post. Rationalists are famously not as successful as they think they should be, and I think that's because they often confuse logic for rationality (even though they know in theory the two are not the same). Sometimes vibing, immersion, practice are the dominant strategies. The people getting millions of views on TikTok are not highly logical, deeply analytical people, and you shouldn't be either most of the time if you want to get lots of views. Put simply, if you want to dress well, you can either study the psychology and history of fashion, or you can just look at 1000 pictures of well dressed people until you absorb the pattern. Vibe your way to success.