AI and Watermarks: The Next Chapter in Regulatory Measures
Leading artificial intelligence creators have agreed to stamp AI-generated content with watermarks, following a meeting at the White House.
Leaders of major tech companies, such as OpenAI and Google, arrived at this decision after consulting with the American President and his team. Despite fierce rivalry, they promised to meticulously scrutinize language models prior to their official application launch and share beneficial information among themselves.
“These commitments are a promising step but we have a lot more work to do together.” - © Joe Biden, the 46th President of the United States.
The maneuverings of American politicians can be viewed as a successful endeavor to create preliminary rules for artificial intelligence. This initiative also earned the backing of other power representatives, including members of Congress and the Senate, since the US is somewhat trailing behind European countries in terms of regulation.
Why Implement Watermarks on AI-Generated Content?
Artificial intelligence has seen a dramatic surge in popularity over the past year. Many people have started to use AI applications to craft both entertaining and work-related content, resulting in outputs that bear the semblance of authentic artistry. This trend has given rise to complex issues related to copyright infringement and AI hallucinations.
The introduction of unique symbols to identify content generated by artificial intelligence, which includes video, text, and audio, may offer a solution to these problems. It allows users to discern whether a piece of content is human-made or a product of machine operation. Moreover, this should curb the production of fake news or defamatory materials.
Will Such Tagging be Effective?
Theoretically, the idea holds merit. However, companies have not disclosed the technical specifics of the tagging process. This is particularly pertinent for text or music formats, which are more challenging to watermark than visual content. The existence of neural networks capable of entirely erasing any markers presents another problem.
How do these companies plan on adhering to their pledges after distributing the content to users? They likely need to devise a novel labeling solution that eliminates the possibility of circumvention. Alternatively, they may choose to persist with ineffective methods for the sake of compliance. Regardless, it remains to be seen how swiftly legislators and developers can implement these new rules.
