10 ms·
Monolith: The Recommendation System Behind TikTok
- cwillu 4y agoReferenced paper at https://arxiv.org/pdf/2209.07663.pdf https://arxiv.org/pdf/2209.07663.pdf
- SpaceManNabs 4y agoSimilar article posted here: https://news.ycombinator.com/item?id=34836877 https://news.ycombinator.com/item?id=34836877
- dang 4y agoThanks! Macroexpanded: The secret sauce of TikTok’s recommendations - https://news.ycombinator.com/item?id=34836877 https://news.ycombinator.com/item?id=34836877 - Feb 2023 (138 comments)
- chadrs 4y agoI must be rare in that the longer I use tiktok, the less relevant the recommendations feel. Maybe because I compulsively watch videos until the end even if I don't like them.
- realfeel78 4y agoSounds like a you problem? At any rate, they added an option to reset it anytime a few weeks ago: https://techcrunch.com/2023/03/16/tiktoks-new-feature-lets-you-refresh-your-for-you-feed-and-retrain-your-algorithm/ https://techcrunch.com/2023/03/16/tiktoks-new-feature-lets-y...
- MuffinFlavored 4y agoI want to know what portion of the algorithm is responsible for, when you are given a new blank slate user, "tries" certain categories like, let's present this user travel or cooking material, that's usually safe then, let's try things like certain genres of music, we'll see what they like/don't like what i don't get is... how does that first recommendation on the #foryoupage or discover or whatever it's called, starts recommending you the sex workers who try to post as close to NSFW material as possible, get you to land on their profile, in their bio is a link to their Instagram or Linktree, and then from there it's an OnlyFans link does the system try to recommend a soft entry into this content and then just pivot away if the user doesn't like it?
- bluefirebrand 4y agoIt probably just recommends a bunch of stuff that is popular at the moment for new users. Or it tries to match you to an existing profile it has from some ad network data or something.
- MuffinFlavored 4y ago> It probably just recommends a bunch of stuff that is popular at the moment for new users. I get that, but I feel like it starts with "known safe/neutral" material like cooking/traveling/photography/whatever How can it detect "hey, this person might like if we introduce softcore porn into their timeline"? Like, do they have softcore porn identified on a scale and they introduce the really "safe" stuff and then gradually crank it up? Why are they presenting softcore porn at all? The Apple App Store is cool with that ToC wise?
- bluefirebrand 4y agoI think you're overcomplicating it. It's not trying to start with "safe" stuff, it's not trying to "gently introduce" softcore porn. It's going "This video got a billion views in the last 30 minutes, people must love it, let's keep amplifying it to any account that hasn't explicitly rejected this category of content" Presumably blank slate accounts are treated as open to anything, until people start curating.
- MuffinFlavored 4y agoduring the curation process, how does it start to slowly introduce sex workers? because when I was on TikTok, it was a non-zero amount of the content
- libraryatnight 4y agoAt the risk of going down a rabbit hole for no real reason, I don't use tiktok but when I speak to those that do I've not yet heard this softcore porn/sex worker thing. For example, in my mind, not all ASMR content might lead to sexualized recommendations, but a girl in a bikini top with cat ears doing ASMR might generate both recommendations for ASMR and other more cam-girl like content. So I guess my question is, when you're starting off in tiktok seeing cooking videos, do you trend towards ones that feature 'sexier' hosts? They might not be sex workers to you, but they might be making tiktok think you're interested. Also, what does tiktok know about you to start? What info do you have to give it to start an account?
- jabbany 4y agoSeems like a common problem of recommenders tbh. Like binary search, they're really good at finding local optima quickly, and then are rather bad at getting out of them once they get there.
- dmix 4y ago> Maybe because I compulsively watch videos until the end even if I don't like them. That would definitely do it, basically destroying their most important signal. TikTok is best in class for recommendeding content and I personally haven't see a dip in quality. Aka I never get trashy videos or anything cringe, just a consistent stream of science/tech, local Toronto restaurant reviews, cat videos, etc
- throwaway290 4y agoThe lack of up/downvote is what destroys their signal, I think. They do it to themselves
- dmix 4y ago? Do you have any proof things aren't working out for TikToks algorithm besides a few people whining on HN who probably weren't the target market in the first place? Because they seem to be doing just fine.
- rendaw 4y agoGP didn't say anything about TikTok's algorithm not working out for people in general, but I'll bite. The selling point behind TikTok is that by using the app normally (i.e. watching what you want) TikTok can figure out what stuff you want to watch. But pretty much every suggestion that comes out on HN for people who complain about the algorithm on sites like this is to do unnatural things to "train" the algorithm: be very careful about how long you watch videos, click on videos that you think are relevant even if you don't want to watch them, watch things you're not interested in to prevent falling into a rut, avoid click on things you're curious about if you don't think you want to watch 100s of them to avoid tainting your recommendations, etc. If that's "working" I think I'd agree that TikTok was mostly hype.
- throwaway290 4y agoYep. It's almost like tiktok targets users who are only using system 1. If you just go with lizard brain reactions you don't care about any of that. But once system 2 is in play you start doing "crazy" things like watch a video forensically even if you don't want to be fed similar videos on the regular or skip videos even if they are about things you like, and you need some control over what you watch
- nickthegreek 4y agoThis can happen to me if it gets stuck down a avenue that it thought I was interested in. But the next day or even a few hours later, it seems to correct itself.
- joshu 4y agolong-press and select “not interested” and it will figure it out pretty quickly.
- gammarator 4y agoThis is a critical step to get rid of bad recommendations, and the algorithm seems to treat it as a very strong signal—after 1-a few “not interested” labels you won’t see that content for weeks if ever.
- thomasahle 4y agoThey are probably trying to save GPU power on already hooked users. This is a common trick in Recommendation Systems. You want to spend the most resources / run your most expensive model on users that are just checking out your platform. A bit like how Poker sites give you better cards in the beginning.
- osti 4y ago>A bit like how Poker sites give you better cards in the beginning. As a former online poker pro, this is probably the most idiotic thing I've read on this site.
- brookst 4y agoEh, it’s bad, but not as bad as the “companies intentionally leak false information to make their upcoming products look unappealing just so the real announcement is a pleasant surprise” 8D chess players.
- HWR_14 4y agoIt's less idiotic if you assume the bias goes the other way. People who stick with online poker had better cards in the beginning leading to them early winning and continuing to play. They then experience a reversion to the mean. Those who lose too often too early lose interest and sign off. They say the worst result to get on your first roulette spin is to hit your number. Obviously, these are just trends among people. Some gambling addicts never won and skilled poker players won't go on tilt after 20 bad hands.
- winternett 4y agoThat's because on the inverse of algorithm recommendations, everyone forgets that the platform is primarily built around an ad system that generates money... There is always going to be a conflict with recommendations because they need to placate and accommodate paying users of all kinds, even users with low quality content and outright commercials. Pretty much all of these social platforms work in this manner now. As a result everyone is seeing undesired (ad) boosted content on top of the other ads (that are marked as ads).
- maxk42 4y agoFollowing the link to the cuckoo hasing algorithm on Wikipedia, I don't quite understand what it's doing. I looked up another couple articles but still find myself confused. Does anyone have a link to a resource with an easy-to-follow writeup of how cuckoo hashing works?
- bhawks 4y agoI found this pretty clear and it includes an interactive visualization: https://www.lkozma.net/cuckoo_hashing_visualization/ https://www.lkozma.net/cuckoo_hashing_visualization/
- CBLT 4y agoI understood it from this HN comment: https://news.ycombinator.com/item?id=8491456 https://news.ycombinator.com/item?id=8491456
- lelandfe 4y agoRecommendation: once you do grok it, see if you can update the intro on the Wikipedia article to help others out :)
- Groxx 4y agoHash collisions happen. What do you do with them? Chained (aka open hashing) hashtables store pointers rather than values in the "main" array, and just put collisions in a linked list on that hash cell. Easy, but indirections have a cost. Closed hashing (aka open addressing, yes it's confusing) hashtables take hash(input) and just do it again to get a second location, and put it there. Repeat N times for N hash(hash(hash(...))) collisions. Dense, but needs more complex logic to figure out when to stop looking / what to do when deleting because anything could be at location X due to a collision with something else. Cuckoo hashtables use two (or more) hash algorithms rather than one, and dedicate a portion of the memory to each algorithm. If something's already in the first algorithm's location, put the thing it's colliding with in that thing's second location. On read, check both locations. Dense, relatively simple for both insert and deletion, and tolerant of a few collisions with low cost. (cuckoo hashtables are a form of closed hashing / open addressing, because they keep all the data within the data-sized arrays, not storing extra info. and all of these are over-generalizing / there are fairly different-looking strategies available, e.g. it's not necessarily pointers or strictly repeated hashing) And you could just reject the existence of collisions entirely and move to a new, larger array immediately. That tends to perform so poorly in both cpu and memory that nothing really does it in practice, but it is technically an option.
- jayzhan 4y agoWhy not linear probing? Cuckoo hashing is usually slower.
- replwoacause 4y agoI've avoided downloading TikTok due to security concerns and its ties to the People's Republic of China. Are these concerns valid? I remember reading articles that claimed it was stealing data from people's phones and spying.
- galacticaactual 4y agoYes.
- replwoacause 4y agoInteresting, and despite these concerns about privacy and security, many high-profile people including celebrities are using TikTok. It's likely they have dedicated devices for social media rather than installing TikTok onto their primary devices but it still amazes me how many people, including important ones, have ignored these warnings. Last I heard TikTok has 1 billion monthly users.
- 6r8346474 4y agoLots of people also just don't care about either China or the data privacy concerns because they're not data private people nor data privacy activists. I don't think it really has to do with having the luxury of a second device so much as the issue not actually being an issue to most people.
- qzx_pierri 4y agoIt’s the only app/service that makes people smile non-stop. It’s endless videos of cool stuff, comedy, cat videos, tech tidbits, cooking videos, etc. I’ve deleted all other social media accounts because browsing Facebook, Instagram, and Twitter made me feel like shit. TikTok doesn’t make me feel that way, and it’s that simple. This is the same reason why a ton of people are dialing back their Instagram usage or creating “finsta” accounts - Who wants to feel insecure constantly? Same goes with Facebook. There’s the argument “Well you’re not using those service correctly. I use FB/IG/Twitter and I’m fine. Follow these steps on how to avoid toxicity on your FB feed…” (articles like this used to be posted to HN constantly). But the thing is… why should we have to do that? TikTok shows me content that makes me smile BY DEFAULT. I am not being milked for engagement using my fight or flight response like Meta’s lolcow. If other apps weren’t so toxic (by design), then people wouldn’t be nervous when posting on IG, or avoiding Facebook because of the cancer comment sections and ranking systems/algorithms designed to make you angry.
- revskill 4y agoI try registered an account on Tiktok once, try uploaded a video of mine, and never get any view. The ideal social network, to me, is everybody has a fair chance to get enough views for them to seriously produce good content to keep followers retention. That means, "more money to get more views" technigue should be obsolete in this system. Tiktok failed at least to me.
- philosopher1234 4y agoYou formed very large conclusions from one attempt
- asdfman123 4y ago> It’s downright addictive. Digital nicotine
- Irishsteve 4y agoThis isn’t the recommendation algorithm or system for tiktok it’s for 2 https://www.byteplus.com/en/product/recommend https://www.byteplus.com/en/product/recommend and it’s outlined in the paper. People miss this that skim it.