[chvid]

The point is not media format (short video clips) but the "sorting hat". That it breaks the traditional social media model with a news feed, discussions, a social graph of friends.

I think it is completely meaningful to call it next generation (or second generation where the social graph model is the first generation).

Also notice that this model can be applied to other media formats: Text, pictures, audio ...

[hardtke]

This thread is completely missing the draw of TikTok, at least as it pertains to my teenagers. TikTok makes it fun and compelling to create content. My teenagers will actually go places with their friends to "make TikToks." The reason it is compelling, and this is not true of any other current social network, is that good content will get eyeballs -- the algorithm seems to be "fair" in terms of playing your content to enough people to see if it is any good, at which point it can get widely distributed. Most social networks start out this way, but eventually good user generated content gets drowned out by influencers and commercial interests. It remains to be seen if TikTok can stay this way.

[creeble]

I wonder how much content there would be if they hadn't licensed music for people to dance / lipsync to.

I haven't seen that many TikTok videos, but I haven't seen many that didn't have music in them, and I don't think any of it was original.

Licensing music has never been easy, and I don't get how two guys (even from silicon Valley) were able to do it from launch.

[newen]

Yep, I’ve come across a couple of times in my for you page videos posted with 0 likes from 0 follower, 0 following accounts. And plenty of low likes and followers accounts. It gives anybody a chance and that makes it engaging.

[prionassembly]

Yes. Even the vaunted algorithm is merely an improved way of doing this. I remember when "the long tail" became a slogan. What was meant was that if popularity is a power law, then at every level there's a shorter long tail. The point was to accentuate this, to find the celebrities in each niche and monetize them. The underlying mental model has always been the one from broadcasting.

Maybe TikTok comes from China because Communist ideology still influences the Chinese; or because they didn't have a Dick Cavett and a Frank Sinatra, celebrity TV. The ceremonies for the 1980 Moscow Olympics had no celebrities, but a diorama of the dozen Soviet cultures from the Ukraine to Kirghistan. The 1984 Olympics in the US had Lionel Ritchie. But Communists or not, the early promise of the internet was that you could participate, and it doesn't feel you can participate on Twitter.

[treis]

>I think it is completely meaningful to call it next generation (or second generation where the social graph model is the first generation).

I don't really see it. We've seen the TikTok model before in Imgur, StumbleUpon, YouTube, Reddit, Twitch, Digg, and probably others. It's mostly memes, funny videos, how-tos, and attractive women. They've hit a sweet spot of editing tools and enforced short format to provide a constant stream of quick entertainment. But I don't see anything earth shattering or ground breaking there.

[villahousut]

In all of the examples you provided there's either no personalization or you have to manually curate your feed. Tiktok and youtube are the only ones where your feed is automatically curate for you.

[Slartie]

> Also notice that this model can be applied to other media formats: Text, pictures, audio ...

Actually, I don't think that this is as easy as you might think. The article goes into this a bit when saying that short video sequences are well-suited for such an algorithm because they provide a high frequency of "inputs" per time unit, but I think the article falls short of describing the other thing that makes videos particularly suitable (and, by extension, makes the assumption that "the TikTok algorithm" had a great future in many other places too, of which I am a bit more skeptical). This other critical thing is that video sequences in general also allow a huge variety of inputs to be gathered from consumption that text, pictures and audio can’t match.

- It is trivial to find out which part of a video a user has seen. This is nearly impossible to do reliably with textual content (assuming you don't have an eye tracker running).

- Instead of a still picture, a video provides much more things for the viewer to see. So instead of just knowing that in a picture there's a cat and you thus deduce the user likes cats, it's basically possible to split a video up in slices of which you know where there's a cat, and where there's a dog, and where there's whatever else, so from just that single video you might deduce info about the users' interest in cat/dog/whatever content all at once (depending on which parts a viewer has seen, which parts were skipped, at which point the viewer aborted, or at which point the like button was tapped).

- Video mostly also delivers audio, hence everything that you can gather from audio, like whether a user tends to prefer female or male voices, or which music style someone prefers, comes as a bonus when gathering info from video viewing

- If your videos' audio features someone speaking some text, you can speech-to-text that content and pump it into the usual machine learning modules, from simple sentiment analysis over trying to determine the topic someone talks about up to full-blown "trying to understand what this person is actually trying to say" and take that as an input for determining a viewers' interests. This is basically text analysis, so it lends itself to textual content as well, and audio too, but not so much to pictures.

Video is just really pumping out the maximum of all of these content formats in terms of potentially relevant data points about someones' interests, and it does so at really high frequency, especially if the length of each video is as short as on TikTok and thus the content producers have already performed the daunting work of condensing lots of content into the least number of seconds possible.

[mistermann]

I'd like to see someone write a blog post taking a shot at (speculatively) "reverse engineering" how the TikTok algorithm works (or may work)...like what attributes it might extract from a video (some of which you've mentioned above) and what it might do with them. Basically, how the overall thing may work, as well as how it may improve over time, taking into consideration current cutting edge ML techniques and speculative future capabilities.

[gruez]

>That it breaks the traditional social media model with a news feed, discussions, a social graph of friends.

don't most social networks have the "sorting hat" in the form of algorithmic feeds?

[nemothekid]

The article goes into this - for people just jumping into the comments:

The author states that while western algorithms are based on your follow graph (e.g. Instagram is relatively useless until you follow someone and even then your feed is based on your follower graph, like what people you follow like), TikTok builds this data on video features. This increases TikTok's stickyness because you don't need to do anything other than use the app for it adjust to your tastes. There's no need to "import" your contacts or suggest people you should follow, it just "knows" after you watch a couple videos.

[mrkramer]

YouTube mastered this years ago and that's why it is 3rd most visited website in the world and 2nd most used search engine in the world.

I see TikTok as a better version of Vine but I still can't understand if TikTok is so much popular and so much worth why did Twitter shut down Vine? Twitter is like modern MySpace it will fail sooner or later if management doesn't get replaced and if they don't start thinking long term.

[9nGQluzmnq3M]

I don't know about you, but my YouTube recommendations tend to be pretty useless. I can't remember the last time autoplay found me genuinely compelling content that wasn't already in my subscribed channels.

[catalogia]

Youtube has been doing this for years. Clear your cookies then start watching youtube videos. Youtube will immediately begin tuning video suggestions to what you watch, no contact importing or channel subscribing necessary.

[disgruntledphd2]

Spotify too.

I'll never forget putting one song into Spotify and having it recap my late 90's listening habits over the course of an evening.

[esrauch]

That sounds just like YouTube and how I presumed Vine worked though? E.g. that the majority of users don't set up any follow graph, and that most content users view is algorithmically-discovered and not like Twitter, Facebook, Instagram where most of the content is based on an explicit graph?

[casefields]

Sounds like the algo's that the porn companies have started movie to. Stickyness works.

[chvid]

Try downloading it and just play with it for a bit without signing up.

Compare that with Instagram.

[mrkramer]

Well on desktop and even on mobile you can't browse Instagram freely without signing up or singing in, it is kinda double edge sword which forces you to join or backs you off.

[Ericson2314]

Did Vine not work that way? I have no idea and the author didn't say either.