Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ByteDance's Recommendation System

github.com

Post reply on HN
Print view View original

51–58 of 58 posts

Re: ByteDance's Recommendation System

#51

Post by solarkraft »

mrfox321 wrote: post #9
xnx wrote: post #7

Like Facebook before, everyone talks about the TikTok algorithm being some super secret and valuable mystery. In reality, both Facebook and TikTok succeeded because they were at the right place at the right time and didn't screw things up. The TikTok recommendation system is smart and very well implemented, but nothing novel that couldn't be implemented by a dozen other teams.

Not true. Why do IG reels and YouTube shorts suck, then? They clearly built something superior. And it can't seem to be matched by the biggest tech companies.

I haven’t used TikTok, but I find both Reels and Shorts addictive enough.

Re: ByteDance's Recommendation System

#52

Post by leobg »

xnx wrote: post #38
newsclues wrote: post #15

Earlier quoted context omitted.

I question the assertion that TikTok has more interaction data than Google.

Google has a ton of interaction data to be sure, but the app design decisions of TikTok (auto play, auto loop, easy swipe, easy like, etc.) extract so much more usable/actionable interaction data. The size of the like button on YouTube is a tiny percent of the screen. On TikTok the like button is the whole video.

Not just that. The whole UI is designed for behavioral data aggregation.

It’s not just “did you click the like button”. It’s “did you swipe it away? How long did you watch until you swiped it away? Did you come back afterwards? Did you let it loop multiple times before moving on?”.

They’ll capture likes and dislikes you yourself probably didn’t even knew you had, just from tens and hundreds of these micro actions. And they’ll do it in the very first hour of you using the app, whereas YouTube won’t know too much about you even after months of you using it.

Re: ByteDance's Recommendation System

#53

Post by yunohn »

PittleyDunkin wrote: post #26
yunohn wrote: post #6

Earlier quoted context omitted.

This is true for all social media algorithms. None of them are purely automated and for good reason. You need humans going in and tweaking the outcomes to ensure users have a good experience. Of course, when the conversation is about TikTok, this often becomes accusations of propaganda. But YouTube, Facebook, and Twitter all exert significant control over their algorithms and things like their Homepage, Trending Topi…

> This is true for all social media algorithms. None of them are purely automated and for good reason. Purely chronological sorting of posts works the best and is fully automated. It's just really bad at keeping users eyeballs on ads.

I really don’t agree with this - chronological sorting favors larger/power users who spam their content through all hours of the day, versus smaller users that you probably care more about.

Re: ByteDance's Recommendation System

#54

Post by yunohn »

dylan604 wrote: post #22
yunohn wrote: post #6

Earlier quoted context omitted.

This is true for all social media algorithms. None of them are purely automated and for good reason. You need humans going in and tweaking the outcomes to ensure users have a good experience. Of course, when the conversation is about TikTok, this often becomes accusations of propaganda. But YouTube, Facebook, and Twitter all exert significant control over their algorithms and things like their Homepage, Trending Topi…

If you don't have a way to manually push the algo, then you'd never be able to sell features like promoted posts and the like. And why would you not want a feature to sell?

Ads can be inserted into any kind of feed, there is no difference between chronological and algorithmic. It’s usually a simple calculation of target ads displayed per posts viewed, with per user/advertiser caps.

Re: ByteDance's Recommendation System

#55

Post by yunohn »

pavlov wrote: post #18
yunohn wrote: post #6

Earlier quoted context omitted.

This is true for all social media algorithms. None of them are purely automated and for good reason. You need humans going in and tweaking the outcomes to ensure users have a good experience. Of course, when the conversation is about TikTok, this often becomes accusations of propaganda. But YouTube, Facebook, and Twitter all exert significant control over their algorithms and things like their Homepage, Trending Topi…

Sure. HN is very actively moderated, and most people here probably agree that it’s worth it. (Those who don’t like it presumably don’t stay here.) But at the massive scale of Meta or ByteDance, there is a difference between removing problematic content and actively promoting content. They’re two sides of the same coin, but the first is applied based on reactive guidelines (“we’ve previously decided this kind of conte…

> HN is very actively moderated

I explicitly did not mention HackerNews, as the homepage feed is primarily based on user voting - neither algorithms nor chronology. Dang’s moderation is not comparable to other social media platform’s feed curation.

> there is a difference between removing problematic content and actively promoting content

Again, there is sufficient evidence that all major social media platforms do exactly this, not just TikTok. Hence why I said:

>> The conservative right often labels such curation as liberal propaganda.

> where this kind of thing is possible and leaves little paper trail

Could you point to the paper trail that Meta, Google, Twitter provide on their curation actions? Otherwise, this just proves my point that people blindly want to accuse Chinese platforms of shady activities, and Western ones as paragons of virtue.

Re: ByteDance's Recommendation System

#56

Post by dylan604 »

yunohn wrote: post #54
dylan604 wrote: post #22

Earlier quoted context omitted.

If you don't have a way to manually push the algo, then you'd never be able to sell features like promoted posts and the like. And why would you not want a feature to sell?

Ads can be inserted into any kind of feed, there is no difference between chronological and algorithmic. It’s usually a simple calculation of target ads displayed per posts viewed, with per user/advertiser caps.

nobody said anything about ads

Re: ByteDance's Recommendation System

#57

Post by yunohn »

dylan604 wrote: post #56
yunohn wrote: post #54

Earlier quoted context omitted.

Ads can be inserted into any kind of feed, there is no difference between chronological and algorithmic. It’s usually a simple calculation of target ads displayed per posts viewed, with per user/advertiser caps.

nobody said anything about ads

What do you think a promoted post is? It’s an ad based on “user generated content”.

Re: ByteDance's Recommendation System

#58

Post by tw1984 »

voxic11 wrote: post #40
tw1984 wrote: post #20

for those interested, Chinese laws forbid the export of recommendation systems, unless ByteDance is challenging the Chinese laws here, which is highly unlikely, this simple can not be the recommendation system used in their production. will be far more interesting to know say what is the difference, what got changed/removed to make them feel comfortable that such an open source variant won't get them into troubles wi…

This is essentially the framework for executing their recommendation system but the actual piece which determines the recommendation is a model called "demo" so I presume its not the actual ML model they use in production.

thank for the details, very useful.
Post reply on HN