Clickbait directly from Google: Test with AI-generated titles in Discover feed
Google's Discover feed is popular for finding worthwhile articles or news on the internet. In an experiment, Google is having titles generated by AI.
(Image: Framalicious/Shutterstock.com)
As part of an experiment, Google is replacing the titles of news articles in the Discover collection in the USA using AI technology – and in the process, producing sometimes hair-raising errors. The US magazine The Verge has noticed this, after which Google has confirmed the attempts. The primary goal appears to be to shorten longer headlines at all costs, even if they become incorrect as a result. For example, The Verge reports that a title created by this, stating that Valve's Steam Machine looks like a console but won't be as cheap, has received the completely wrong title: "Steam Machine price revealed". An article about Mindfactory selling an AMD graphics card more frequently than all of Nvidia's cards received the completely misleading title from Discover: "AMD beats Nvidia".
Dangerous for the media
In Discover, Google collects articles from various online media, thus seemingly offering an overview of the latest news and worthwhile texts, regardless of their source. In the Google app or on the mobile website, users can scroll through them, and with a click, they land directly on the respective news pages. For media outlets, this is a welcome source of visits, but in most cases only if interested parties are not lured in by sensational or completely false headlines. Because those who feel so misled are likely to avoid the respective news page or Google Discover altogether in the future.
The observations by The Verge suggest that the focus of the AI used is on shortening longer titles. Each of the examples listed consists of exactly four words. Not all of them are incorrect or misleading; some just sound much less interesting than the original. For instance, an article about a boy's award-winning origami model that can hold ten thousand times its own weight was reportedly given the title "Origami wins prize". Others, in this forced brevity and without context, simply no longer make sense, explains the US portal. You can only see that the titles are AI-generated when you click on the corresponding article preview. However, it does not state that Google is responsible for the AI title.
Videos by heise
The titles replaced by a Google AI were carefully formulated by the authors, The Verge reminds us. The goal is, of course, to generate interest in the article, but for many media outlets, it is also explicitly about not creating false expectations. If Google itself now produces this so-called "clickbait" and the media outlets have no way to prevent it, their reputation can suffer. Google has now assured The Verge that the experiment is only visible to a small portion of Discover users. However, it was not only The Verge that noticed this, but also Engadget. Google has explained that the AI titles are intended to make topics "more digestible" before links from the internet are explored.
(mho)
