In 2012 Kevin Systrom and Mike Krieger sold their app to Facebook for around one billion dollars. It was called Instagram, had been founded two years earlier, and at the time was still known mainly for its photo filters. Stories (copied from Snapchat), reels and, above all, the one billion four hundred million users who today visit the platform — now one of the most valuable assets in the Meta group — were still a very long way off. After selling the app, the two took a lengthy period of reflection, collaborated on a few projects but avoided putting their names to a new product. Until last week, when they introduced Artifact, a social network conceived as a "TikTok for text-based content", which is currently available to a limited number of people (you simply need to register on the website).

What Artifact is, the new social platform from the "fathers" of Instagram

Artifact aims to be "a personalised news feed governed by an artificial intelligence", harnessing recent advances in machine learning to deliver users a better experience — one tailored specifically to them. According to what Systrom explained to The Verge, the two have been working on it for a couple of years. They waited so long because they had promised themselves not to found a new company until three events had occurred: first, a major breakthrough in the field of consumer technology (in this case, AI and the TikTok phenomenon); then the emergence of a new way of connecting this breakthrough to the world of social media, which they consider their reference sector; and finally, a clear and precise feature. In short, Artifact was born to solve a problem.

To do so, it borrows a piece of the language model Gpt-3 (the one behind ChatGpt and Dall-E). Specifically, the "T", which stands for "transformer" and denotes the ability of these systems to understand human language more easily and with greater precision, guaranteeing the sector a faster rate of evolution. More than to OpenAI, this particular technology is owed to Google, which invented it in 2017. Beyond generating texts and images, all of this can also have positive repercussions for social networks, overturning their basic premise: instead of showing you content from your friends (the Facebook model) or from accounts your friends follow (Twitter), TikTok began showing external content on the basis of other parameters that had nothing to do with users' "social network". A Copernican revolution that Artifact seeks to apply to text-based content as well. The co-founders stress that they want to work closely with publishers to ensure a balanced and sustainable information offering (for instance by displaying advertising from news sites), and that they see text as the first phase of a broader project through which they want to refine their AI.

The problem of disinformation

The idea of a news feed highly personalised by an algorithm, however, also brings to mind some rather unpleasant precedents, such as those that affected Facebook in particular from 2015 onwards and that led to the many controversies and scandals surrounding the so-called filter bubbles, fake news and deepfakes. Problems that also concern, albeit in a different way, TikTok itself, whose algorithm tends to distribute content that is always in line with users' existing positions. The risk of falling down the rabbit hole and into a path of online radicalisation therefore still exists, and it is not yet clear how much artificial intelligence can help to avoid it.