Code and Dagger

View Original

Military Wants an AI to Watch You Work, Correct Your Mistakes

(Photo by Rock'n Roll Monkey on Unsplash)

In news that certainly won’t induce nightmares of a near-omniscient and overbearing Clippy, the mad scientists of the U.S. military are seeking to develop an artificial intelligence that would watch military personnel as they work and correct their mistakes as they make them.

Dr. Bruce Draper, a program manager in the Defense Advanced Research Projects Agency (DARPA), said that in the “not too distant” future service members decked out with various sensors, from microphones to augmented reality headsets, could be “assisted” in whatever work they’re doing by a vastly more involved version of a virtual assistant.

“What we need in the middle is an assistant that can recognize what you are doing as you start a task, has the prerequisite know-how to accomplish that task, can provide step-by-step guidance, and can alert you to any mistakes you’re making,” Draper said in a write-up on the proposed tech on DARPA’s website.

To get to this dreamscape (or hellscape depending on your point of view), DARPA established the Perceptually-enabled Task Guidance (PTG) program “to explore the development of methods, techniques, and technology for AI assistants capable of helping users perform complex physical tasks,” according to DARPA’s press release.

“To develop these technologies, PTG seeks to exploit recent advances in deep learning for video and speech analysis, automated reasoning for task and/or plan monitoring, and augmented reality for human-computer interfaces,” the website says.

For a start the program wants to see how such a technology could be applied to tasks like “mechanical repair, battlefield medicine, and/or pilot guidance.”

Creating such an assistant is hardly simple, and DARPA says the research has been split into component areas to deal with specific problems — like how to make the AI recognize any external developments that might change the way something has to be done, while ignoring stimuli that’s unrelated to the task at hand.

Then there’s the question of how often to annoy the user — and keeping an eye on how annoyed the user might be at any given time.

“PTG assistants must be able to determine how much information to share with a user and when to do so,” DARPA says. “This requires developing and integrating an epistemic model for what the user knows, a physical model for what the user is doing, and a model of their attentional and emotional states.”

The write-up on the DARPA was in part to announce an upcoming “Proposers Day” in which outside groups will be able to learn more about how they might throw their hat in the ring to help develop the technology and win related military contracts.

Whatever comes of the tech, the world has come a long way from Clippy, the office assistant that even Microsoft called “infamous" when the animated paper clip was discontinued in 2001 (only to be resurrected and killed off again in 2019).

Siri, for instance, is on millions of Apple phones with no signs of going anywhere. That could offer some indication of the seriousness of DARPA’s mission — after all, Siri was developed using DARPA tech.

[Do you have a tip or question for Code and Dagger? Send it along at CodeAndDagger@protonmail.com. And if you like what you read and want to help keep the site running (kind of) smoothly, click here to learn how you can lend your support. ]