Twitter has announced that it will take over Fabula AI, hoping to be able to fight better with fake news. Fabula AI is a startup from London with a deep learning platform that detects disinformation on online platforms.
However, Fabula AI has not yet had time to make its platform generally available, writes Silicon Angle. The startup wanted to do that later this year.
Fabula AI’s software uses an emerging technique for artificial intelligence (AI) called geometric deep learning. On this basis, the platform searches for patterns in online content that can reveal the existence of disinformation. The startup algorithms do not analyse the content itself, but evaluate the way and speed at which it is distributed to determine whether it is suspicious.
The AI therefore does not need to understand exactly what is in a message, making it easy to apply to different languages, without the need for major changes. This makes it easy for Twitter to implement in its own technology.
The decision of the social medium to buy Fabula AI does not only have something to do with the technology of the startup. The history of the team that set up the startup also plays a role.
The chief scientist, for example, is Michael Bronstein, a professor at Imperial College in London and the chairman of machine learning and pattern recognition at the computer science department at the university. Federico Monti is the chief technologist, and a doctoral candidate doing research on geometric deep learning.
The intention is for Bronstein to become part of Twitter and take on the function of head of graph deep learning. He and the rest of the Fabula AI team become part of the Cortex machine learning group of Twitter. The amount of money involved in the takeover is unknown.This news article was automatically translated from Dutch to give Techzine.eu a head start. All news articles after September 1, 2019 are written in native English and NOT translated. All our background stories are written in native English as well. For more information read our launch article.