Science & Technology

Instagram and Facebook must mark phoney AI photos

 

Facebook and Instagram will implement a labelling system for any artificial intelligence-generated photographs deemed false.

The deployment will occur on Facebook, Instagram, and Threads platforms.

Meta currently applies categorization to artificial intelligence-generated photos produced by its algorithms. The statement expresses the expectation that the next technology, now under development, would generate significant progress in addressing the issue of AI fakery, hence propelling the industry forward.

Meta, as stated in a blog authored by senior executive Sir Nick Clegg, plans to increase its identification of artificial intelligence counterfeits shortly.

During an interview with the Reuters news agency, he acknowledged that the technology was still developing. Still, he expressed the company's intention to create a feeling of progress and motivation for other companies to emulate.

'Simple to avoid'

However, Prof Soheil Feizi, the head of the Reliable AI Lab at the University of Maryland, proposed that circumventing such a system may be straightforward.

"According to the interviewee, it is possible for them to train their detector to accurately identify and highlight images that are specifically generated by certain models," he informed the BBC.

"However, these detectors can be easily circumvented through the use of lightweight image processing techniques, and they also tend to produce a significant number of incorrect identifications." A wide variety of applications are not feasible.

Meta has recognized that its tool cannot process audio and video, the primary media types that raise concerns about AI-generated fakes.

The company has said that it now requests users to categorize their audio and video submissions and that it "may impose sanctions if they neglect to comply."

Sir Nick Clegg has acknowledged the impracticability of verifying text that technologies like ChatGPT have produced.

"The opportunity has been missed," he said to Reuters.

 

'Inconsistent' media policy

The Oversight Board of Meta issued a critique on Monday, denouncing the company's policy on altered media as "incoherent, devoid of compelling justification, and improperly fixated on the manner in which content is produced."

Meta financially supports the Oversight Board but operates autonomously from the corporation.

The critique was prompted by a verdict on a film featuring US President Joe Biden. The video in issue manipulated pre-existing footage of the president with his granddaughter to create the illusion that he was engaging in unlawful physical contact with her.

Since the video was not altered using artificial intelligence and portrayed Mr Biden engaging in behaviour that he did not exhibit, rather than making false statements, it did not infringe on Meta's policy on modified material. Consequently, the video was kept.

The Board concurred that the film did not violate Meta's existing regulations about fabricated material, although it emphasized the need to update those regulations.

Sir Nick told Reuters that he broadly agreed with the ruling.

He acknowledged that Meta's current approach needs to be revised for an environment with a much higher amount of synthetic and hybrid material compared to previously.

Since January, the corporation has implemented a policy requiring political advertisements to indicate the use of digitally manipulated photos or videos.