The United States Air Force and Space Force have introduced a new generative artificial intelligence system called "NIPRGPT," designed to enhance information acquisition and drive departmental modernization. This ChatGPT-like AI tool will assist personnel in conducting secure and efficient online communication, learning, and coding tasks.
According to senior officials of the Air Force Department, although NIPRGPT is still under development, it has already achieved positive results and is considered a "remarkable" technological breakthrough. It will be a significant boost for the Air Force Department's development of skills and flexibility, being part of several modernization initiatives.
Image source note: The image was generated by AI, and the image authorization service provider is Midjourney
NIPRGPT will provide the best tool support for teams after commercial AI solutions have passed security reviews. With the introduction of more complex features, it will offer personnel more opportunities to experiment and learn, and evaluate the performance of different AI models.
The Air Force Department plans to collaborate with commercial partners to assess the specific needs of the forces for generative AI, and then select the technology provider. Some analyses suggest that different AI models may perform differently in specific scenarios.
NIPRGPT originates from the Air Force Research Laboratory Information Directorate's "Dark Sword" program. The program aims to develop next-generation deployable software and AI capabilities to support the U.S. military's operational needs.
Meanwhile, the Defense Advanced Research Projects Agency (DARPA) is also working to eliminate stopgap methods and improve the performance and accuracy of military AI. The agency has launched the "Artificial Intelligence Quantification" (AIQ) project, aimed at examining and understanding AI capabilities, and providing mathematical assurance for the safe and ethical use of autonomous and semi-autonomous technologies.
The AIQ project will focus on specific issues and types of problems, establishing a rigorous foundation for evaluating AI capabilities, and creating model evaluation methods over a three-year period.