The ongoing conflict in Ukraine offers a glimpse of how artificial intelligence is transforming modern warfare. What future advancements in this technology are likely to revolutionize how war is waged?
John Bohannon, Primer’s Senior Director of Data Science, discussed this and the importance of continued research and development in explainability and transparency with Dan Faggella on the AI in Business Podcast. The following is a brief summary capturing select highlights of their discussion.
What AI advances have we seen in the Ukraine conflict?
In examining the conflict in Ukraine, Bohannon and Faggella delve into specific examples of how AI is currently being used in conflict, such as drone strikes and cyberattacks. One key development they discussed is how AI changes symmetric warfare, a conflict between parties with equal access to modern AI tools and technology. This shifts the advantage to the party that can more nimbly apply AI, rather than more traditional warfare capability differences.
Historically, the technologies of war have been developed largely by governments. But now, as Bohannon points out, most of the AI technology relevant to warfare is emerging from the private sector. The drones deployed in Ukraine, for example, are most likely using open source AI software called Yolo.1 The capability to recognize objects such as vehicles and soldiers can be powered by a computer as small as a cell phone.
AI warfare is a game of speed and precision
Bohannon explored how AI text and vision models are advancing military operators’ ability to compete in fast-moving, chaotic battle environments. The following includes a few examples discussed:
- Advancements in AI language skills have gone beyond single-language machine translation. Now, AI models can understand and converse in multiple languages. Military operators can now summarize information from foreign languages in near real-time, even transform the output into standard military report formats. This technology makes it possible to process and extract time-sensitive information from captured enemy laptops immediately on the ground, rather than sending it to a distant base and waiting days or even weeks for the output.
- AI vision technology can help gather and analyze information from low-resolution images collected from disparate sources. Bohannon discussed how military teams might not have eyes in the sky but by using data fusion across both military sensors and open source media. He also discussed the importance of human-in-the-loop expert analysts for validating the insights from AI systems.
“Now it’s possible to just say what you want and the machine can find it for you.”
The future of AI is multi-model
Bohannon delved into the work already being done here in combining vision and text AI models so that we get to a point where our military’s AI behave more like J.A.R.V.I.S.2 Meaning that military operators should be able to converse with their AI and simply request what they require. And this is becoming closer to reality with some of the latest advances in multi-model AI.
“The user interface would look a lot like J.A.R.V.I.S., from the Marvel Universe, Iron Man’s AI sidekick. Because it is the most natural way of interacting with any system … everyone on the ground would have access to this [AI] team that can pore over the data and answer their questions.”
Bohannon warned though that the accuracy and effectiveness of AI is limited by the quality of the data it is trained on. Bohannon also highlights the need for ongoing research and development in the field of AI, particularly in areas such as explainability and transparency. He posits that the more we understand about how AI works and how it can be used, the better equipped we will be to harness its potential and the more trust that users will have in it.
To learn more about Primer’s work with the defense and intelligence communities, visit https://primer.ai/get-mission-ready-with-primer-ai/.