The company already draws attention to content produced by its own systems in this way. It should now also apply to contributions from other parties.
However, as the representative of the company, Nick Clegg, admitted, the general labeling of content from AI can be expected in a few months. In an interview with the BBC, Clegg admitted that the technology needed to recognize artificially created content is not yet at a sufficient level, and certainly will not be for video or audio content.
No social scoring. Artificial intelligence will receive clear rules in the EU
University of Maryland AI expert Soheil Feizi doubts Meta will succeed. According to him, artificial intelligence will always have the upper hand.
“They can train the detector to point to some images specifically generated by some particular models, but such detectors can easily be bypassed with some light post-processing of the image,” Feizi noted.
According to him, there is also a risk of a high error rate in the opposite direction, i.e. that the system will mark authentic content as fake.
The meta has recently faced criticism in both the UK and the US for its computer manipulated content. In one fake video, US President Joe Biden inappropriately touches his own granddaughter.
On the X network, artificially created pornographic images of pop star Taylor Swift have recently surfaced. One of them collected 45 million views and was shared by 24 thousand users.
2024-02-07 11:37:00
#Facebook #tag #images #created #artificial #intelligence #News
