This week in AI and Machine Learning: ML for protecting firefighters, AI-supported therapy, and more.

A Note from the Author

Honeybees play an essential role in the world food supply chain, adding billions of dollars in value to the nation’s agricultural output each year. Saturday is World Honey Bee Day, but celebrate these essential insects a day early with a look at our latest blog.

Artificial Intelligence News

Keeping Firefighters Safer with AI 

Flashovers, a deadly phenomenon in which nearly all combustible objects in a room suddenly catch fire, are among the leading causes of death for firefighters. Historically, firefighters have relied on temperature data to predict flashover risk, but machine learning-based approaches have been introduced more recently. However, even these have had their shortcomings.  

Trained with familiar environments, existing ML-based flashover prediction tools haven’t accounted for the unpredictability inherent to a firefighter’s job. When firefighters storm into a burning building, they’re not just contending with a high-risk, time-sensitive situation, but also attempting to navigate an unfamiliar environment. Firefighters often learn the floor plans of buildings on the fly as they also try to discern the location and severity of blazes. 

Researchers at Hong Kong’s National Institute of Standards of Technology (NIST) recently introduced FlashNet, a model supported by a graph neural network that dramatically improves upon existing approaches. Many members of the team have worked on flashover predicting models in the past and learned firsthand where they tended to come up short. Wai Cheon Tam, one co-author of a new study introducing FlashNet, elaborates, “Our previous model only had to consider four or five rooms in one layout, but when the layout switches and you have 13 or 14 rooms, it can be a nightmare for the model.” Read this ScienceDaily article to learn more about how FlashNet came to predict flashover occurrences with more than 90% accuracy.

Could AI-Supported Help Chronic Pain Patients

A study published in JAMA Internal Medicine has found that a course of cognitive behavioral therapy (CBT) sessions supported by artificial intelligence produced the same or better results in patients with chronic pain as a traditional therapy program. AI-supported therapy required doctors to devote less time to each individual program, potentially making it more accessible to patients. By making pain-focused CBT a more appealing alternative to medication, researchers are hopeful they can reduce reliance on addictive medications and address the ongoing opioid epidemic.

Trials conducted by researchers at the University of Michigan broke 278 chronic pain patients into two groups. The first group conducted ten 45-minute CBT sessions with a therapist over the phone. The second received therapy supported by AI. During automated calls, these patients described their symptoms and the AI “therapist” offered recommendations based on the conversation. Patients were offered either a 45-minute session with a doctor, a 15-minute session with a doctor, or an automated session covering similar subjects.

After three months, patient outcomes were about equal. At the six-month point, AI-supported patients were outpacing the others in terms of improvement. These patients spent longer in the program, received more treatments, and saw better results – all while requiring less clinician time. John Piette, the study’s lead author, hopes to see scientists take a similar approach to making treatment for depression and post-traumatic stress more accessible. Learn more about Piette and his team’s research.

AI’s Healthcare Revolution: Why We’re Still Waiting

“Artificial intelligence was supposed to transform healthcare,” but Ben Leonard and Ruth Reader note, “It hasn’t.” In a new article for Politico, the pair examine both the overambitious goals established for AI in recent years as well as the various legal and technological obstacles that have stymied healthcare innovation. 

Infrastructure, Leonard and Reader argue, presents the biggest challenge to AI’s successful deployment in the healthcare space. Healthcare systems each have unique technological frameworks and many are just now establishing dedicated engineering teams to support their AI goals. What’s more, the government is only beginning to step into its full regulatory role. The Food and Drug Administration is still in the early days of determining how it will evaluate AI solutions and fundamental questions remain about how to best manage technology as it evolves over time. 

Check out some of the ways hospitals are deploying computer vision to monitor activity in real time, optimize processes, conduct audits, and see some of AI’s impressive capabilities.

Join our Community

See you next week! Until then, keep the conversation going on Plainsight’s AI Slack Channel

About the Author & Plainsight

Bennett Glace is a B2B technology content writer and cinephile from Philadelphia. He helps Plainsight in its mission to make vision AI accessible to entire enterprise teams.

Plainsight’s vision AI platform streamlines and optimizes the full computer vision lifecycle. From project strategy, through model deployment, and ongoing monitoring, Plainsight helps customers successfully create and operationalize vision AI applications to solve highly diverse business challenges.

 View All Blogs