NIST Introduces New Platform for Evaluating Generative Artificial Intelligence Capabilities
On Monday, the National Institute of Standards and Technology (NIST), a U.S. Commerce Department agency responsible for developing and testing technology for the government, companies, and the broader public, announced the launch of NIST GenAI. This new program is spearheaded by NIST to assess generative AI technologies, including text- and image-generating AI.
What is NIST GenAI?
NIST GenAI will release benchmarks, help create "content authenticity" detection systems (i.e., deepfake-checking), and encourage the development of software to identify the source of fake or misleading AI-generated information. This program aims to address the growing concern about AI-generated misinformation and disinformation.
The NIST GenAI Program
NIST GenAI will issue a series of challenge problems intended to evaluate and measure the capabilities and limitations of generative AI technologies. These evaluations will be used to identify strategies to promote information integrity and guide the safe and responsible use of digital content.
First Project: Pilot Study on Text-Generated Media
The first project under NIST GenAI is a pilot study to build systems that can reliably distinguish between human-created and AI-generated media, starting with text. While many services claim to detect deepfakes, studies have shown them to be shaky at best, particularly when it comes to text.
How to Participate
NIST GenAI is inviting teams from academia, industry, and research labs to submit either "generators" (AI systems that generate content) or "discriminators" (systems designed to identify AI-generated content). Generators must generate 250-word-or-fewer summaries provided a topic and a set of documents. Discriminators must detect whether a given summary is potentially AI-written.
Data Requirements
To ensure fairness, NIST GenAI will provide the data necessary to test the generators. Systems trained on publicly available data and that don’t comply with applicable laws and regulations won’t be accepted.
Timeline
Registration for the pilot will begin May 1, with the first round of two scheduled to close August 2. Final results from the study are expected to be published in February 2025.
The Growing Concern about AI-Generated Misinformation
The launch of NIST GenAI comes as the volume of AI-generated misinformation and disinformation grows exponentially. According to data from Clarity, a deepfake detection firm, 900% more deepfakes have been created and published this year compared to the same time frame last year.
Alarm and Concern
This is causing alarm, understandably. A recent poll from YouGov found that 85% of Americans are concerned about misleading deepfakes spreading online.
NIST’s Response to President Joe Biden’s Executive Order on AI
The launch of NIST GenAI is part of NIST’s response to President Joe Biden’s executive order on AI, which laid out rules requiring greater transparency from AI companies about how their models work and established a raft of new standards, including for labeling content generated by AI.
Appointment of Paul Christiano to the AI Safety Institute
The launch of NIST GenAI is also the first AI-related announcement from NIST after the appointment of Paul Christiano, a former OpenAI researcher, to the agency’s AI Safety Institute. Christiano was a controversial choice for his "doomerist" views on AI safety and his proposal for an AI alignment initiative.
Conclusion
The launch of NIST GenAI is a significant step in addressing the growing concern about AI-generated misinformation and disinformation. By providing benchmarks, creating content authenticity detection systems, and encouraging the development of software to identify fake or misleading AI-generated information, this program aims to promote information integrity and guide the safe and responsible use of digital content.
Related News
- Fundraising: UK in-home healthcare provider Cera raises $150M to expand its AI platform
- Transportation: CES 2025: Self-driving cars were everywhere, plus other transportation tech trends
- AI: Researchers open source Sky-T1, a "reasoning" AI model that can be trained for less than $450
- AI: Microsoft accuses group of developing tool to abuse its AI service in new lawsuit
