Training Human and AI Teams

Imagine a group of people sitting around a table, solving a difficult problem. Their boss interrupts to introduce a new teammate. The new teammate? She’s a robot.
 

Say what?

Combining human intelligence with the strengths of artificial intelligence (think: autonomous agents) is a concept called Crewed-Uncrewed Teaming.
 

Why are we talking about it?

Crewed-Uncrewed Teaming is a trend shaping how military personnel train for mission readiness – specifically, it’s a process where humans provide feedback to AI to improve its performance over time.
 

Why is it important?

A successful human-AI team requires a level of trust we don’t quite understand… yet.

When it comes to training, experience is built through practice. When it comes to teaming, trust is built through experience. If humans and AI don’t play together on the same team, humans can’t trust AI to do the right thing, and AI won’t receive the right instruction from humans.

Lockheed Martin is using our expertise in connecting disconnected systems to think about training and design together.

How?

We are investing in research and development to train artificial intelligence models across Lockheed Martin.

In one example, we’re training AI agents to help F-35 pilots. By training an AI agent to minimize detection by radar, we’re able to provide student pilots with alternate flight paths they could have taken to reduce detection. Those alternate flight paths are valuable teaching tools to help pilots train more effectively.

This approach will enable us to create more efficient and effective artificial intelligence models that humans can rely on, to shape what the future will look like.