OpenAI, artificial intelligence research lab behind ChatGPT and DALL-E, is taking a step towards combating deepfakes with the release of a new detection tool. This comes amid growing concerns about the potential misuse of AI-generated content in the upcoming elections.
According to a report by The New York Times, the deepfake detection tool, designed to identify images created by OpenAI’s own image generator DALL-E, boasts an impressive 98.8% accuracy rate. However, OpenAI acknowledges it is just one piece of the puzzle. The company is partnering with disinformation researchers to refine the tool and explore additional solutions.
The pressure on the AI industry to address the issue of deepfakes is mounting, especially with a year full of crucial elections globally. Recognizing the limitations of a single-source detector, OpenAI is also collaborating with tech giants like Google and Meta. Together, they are backing the Coalition for Content Provenance and Authenticity (C2PA) which aims to establish a standard for digital content labeling. This “nutrition label” would disclose the creation or alteration process, including AI manipulation. OpenAI is further exploring ways to “watermark” AI-generated audio for easier identification. These watermarks, designed to be tamper-proof, would offer another layer of verification.
In an another development, OpenAI and Microsoft on Tuesday announced launch of a $2 million Societal Resilience Fund to promote AI education and literacy among voters and vulnerable communities. Several organisation including Older Adults Technology Services from AARP, the Coalition for Content Provenance and Authenticity, International Institute for Democracy and Electoral Assistance and Partnership on AI will use the funds to deliver AI education and combat risks of AI and deepfakes ahead of the elections.
“In a year when two billion people across the globe will vote in democratic elections, it is more important than ever to provide tools and information that will help people navigate an increasingly complex digital ecosystem and find authoritative resources,” Microsoft said in a blog.
