“Imagining that users subscribe to receive news is a very challenging model, assumes that people really know what they like”: it is 2013 and Zhang Yiming, who would later become the founder of TikTok, writes this sentence in an article commenting on the closure of Google Reader.
At that time, Zhiming is the CEO of ByteDance, whose flagship product her name is Jinri Toutiao, which means Headlines of the day: it is a particular app for those years, it offers news from the most disparate sources to users without them having to do anything. There is an algorithm that analyzes the behavior of readers, since time they spend on items up to the type of information based on the time of day. In China, Toutiao is a success and it is the infrastructure on which what, a few years later, will become ByteDance’s real flagship product: TikTok.
Social network
What is algospeak, the language used online to escape the TikTok shadow ban
by Emanuele Capone, Francesco Marino
Artifact He Toutiao Western
The history of Toutiao is even more interesting today after that the founders of Instagram, Kevin Systrom and Mike Krieger, unveiled Artifact. It is a platform that has an architecture similar to that of Toutiao and TikTok: a feed of articles from different sources which, after some browsing time, becomes more and more personalizedbased on your preferences and usage habits.
At the moment, you can enter via a waiting list, even if the app doesn’t seem to be available yet in Italy (but you can leave your phone number to get in line). The platform should have 3 functions main: firstly, a default feed entirely managed by the algorithm; then one dedicated to the personal profiles that the user has chosen to follow, where it will be possible to see the users’ comments on the articles; a space to exchange private messages.
Systrom and Krieger’s idea is that Artifact is a space dedicated to verified news. In Casey Newton’s article that broke the news, the two explained that the platform will only include sources capable of adhering to a set of editorial standards to be respected and that it will guarantee access to all political views. They also guarantee an automatic system for recognizing fake news
As you read, Artifact should be optimized starting from read times. In other words, the system will privilege (a bit like TikTok) above all the contents on which the user spends a lot of time. Which is a way to reward articles that are able to really engage readers beyond the title.
Social
The secret of TikTok is not (only) the algorithm
by Andrea Daniele Signorelli
Artificial intelligence and social networks
At the heart of the system is artificial intelligence. And in particular those innovations, first of all neural net Transformerwhich are the basis of many of the latest evolutions in the world of AI, from the improvement of social network recommendation systems, TikTok is tutti, but to ChatGPT.
And it is precisely artificial intelligence that has favored a paradigm shift in the world of social networks: from networks based on social relationships, such as facebook or instagram, to those based on interests, such as TikTok itself or YouTube or the Reels. Systems, such as Artifact, which do not rely on the user’s decisions to select the best contents: they let an automatism choose, based on the habits of fruition.
Systrom explained to Newton that “over the years what I have seen is that whenever machine learning is used to optimize the user experience, things get better very quickly. I watched these changes and thought they might be the future of social media. I’m talking about this seemingly disconnected structure, learned rather than explicitly created by the user. I wondered because it doesn’t happen everywhere in social networksbecause Twitter and Facebook are still based on followers”.
The answer to this requestand therefore the applicability of the TikTok model on articles and news, will say a lot about the future of Artifact.