Deputy CTO, Critical Tech
Department of Defense
DOD to Host Conference on Military Applications of Commercial AI Tech
Maynard Holliday, the Pentagon’s deputy chief technology officer for critical technologies and a past Potomac Officers Club speaker, said the Department of Defense is organizing a conference to seek input from key figures in the industry, academia and the defense sector about how the military can develop trusted artificial intelligence capabilities using commercial technologies.
The three-day conference is expected to take place at the MITRE facility in Virginia beginning June 20, with Heidi Shyu, DOD undersecretary for research and engineering, as host. It will feature presentations from DOD’s Acquisition Undersecretary Bill LaPlante and Chief Data and AI Officer Craig Martell and discussions with AI innovators, Breaking Defense reported.
In an interview with Breaking Defense, Holliday highlighted the importance of having the Pentagon own the technical baseline of military-specific applications of commercial technologies to have control over solution updates instead of relying on a single vendor to evolve a capability. Owning the technical baseline, however, is a challenge to the DOD since private sector players want to protect their intellectual property. Holliday said the DOD can address the IP issue by developing its own dataset containing defense-specific information to ensure the resulting products can be trusted.
The deputy CTO also shared with Breaking Defense that the conference will also include discussions about the pros and cons of generative artificial intelligence like OpenAI’s ChatGPT, mirroring his remarks during a POC event in March. According to Holliday, conference participants will talk about mitigating the tendency of large language models to generate wrong answers to queries, noting that generative AI could serve as decision support if it could be made reliable.
Category: Speaker News
Tags: artificial intelligence Breaking Defense ChatGPT DoD Maynard Holliday Speaker News technical baseline