NIST Launches New Platform to Evaluate Generative AI
NIST has launched a new initiative, NIST GenAI, to effectively assess and manage generative AI technology. This includes developing systems to identify and verify authenticity in digital content. This move is a direct response to growing concern over AI-generated disinformation and it
The US National Institute of Standards and Technology (NIST), part of the Department of Commerce that develops and tests technology for both government and the public, recently announced the launch of NIST GenAI. This new program is focused on evaluating generative AI technologies, including systems that generate text and images.
Evaluation and Guidelines for Authenticity
Through NIST GenAI, the organization will release benchmarks and assist in creating systems to detect "content authenticity," such as systems for checking deepfakes. The program will also promote the development of software that can identify the source of false or misleading information generated by AI. On its newly launched website and in a press release, NIST explains that the GenAI program will initiate a series of challenges to "evaluate and measure the capabilities and limitations of generative AI technology."
First Project Focuses on Media Identification
NIST GenAI's first project is a pilot study aimed at developing systems that can reliably distinguish between content created by humans and content generated by AI, starting with text-based content. Participants from academia, industry, and research labs are invited to contribute either "generators" – AI systems that generate content – or "discriminators," systems designed to identify content created by AI.
A Study With Strict Rules
In the study, the generators must be able to create summaries of up to 250 words from a given topic and document, while the discriminators must determine whether a summary might have been written by AI. To ensure fairness, NIST GenAI will provide the necessary data for testing the generators. Systems trained on publicly available data, or that don't "comply with applicable laws and regulations," will not be accepted.
Join the Testing
Registration for the pilot opens May 1, with the first round scheduled to conclude on August 2. Findings from the study are expected to be published in February 2025.
The launch of NIST GenAI and its focus on deepfake studies comes at a time when the amount of disinformation and misinformation generated by AI is growing exponentially. According to data from Clarity, a company that detects deepfakes, 900% more deepfakes have been created and published this year compared to the same period last year. A recent survey from YouGov shows that 85% of Americans are concerned about the spread of misleading deepfakes online.
NIST GenAI is part of NIST's response to President Joe Biden's executive order on AI, which requires greater transparency from AI companies about how their models work and establishes new standards, including labeling of AI-generated content.