Regard Teams Up With OpenAI to Develop Chatbot Functionality Built On GPT-4

Regard Teams Up With OpenAI to Develop Chatbot Functionality Built On GPT-4


Regard is adding new capabilities to its chief product, an AI co-pilot that helps clinicians diagnose medical conditions. On Monday, the Los Angeles-based startup announced it is teaming up with OpenAI to release new chatbot functionalities built on OpenAI’s large language model GPT-4.

The startup was founded in 2017 by Eli Ben-Joseph, Nate Wilson and Thomas Moulia — they created the startup after they abandoned their plans to go to medical school.

“We all took the MCAT, but all three of us hit the brakes and turned away from that path. For me, personally, I turned away from going into medical school because many of the doctors that I shadowed didn’t seem very happy,” said Ben-Joseph, who serves as Regard’s CEO.

A decade later, the clinical burnout crisis is even more acute. This is due to the fact that healthcare has experienced an explosion of data availability in recent years, but the tools that clinicians are given to take advantage of that data haven’t changed, Ben-Joseph explained. Without workflow-specific automation tools, clinicians are left combing through this data manually.

Regard’s mission is to provide health systems with software that helps doctors and nurses take advantage of all the data in the EHR, Ben-Joseph declared. With this technology, clinicians don’t have to spend the entire patient visit looking at their screen and typing, he pointed out. 

The startup’s AI co-pilot is deeply integrated with both Epic and Cerner, Ben-Joseph said. It analyzes patient data from the EHR — including labs, imaging, clinician notes, histories and vitals — and then generates diagnoses. More than 100 hospitals use Regard’s technology — including Penn Highlands Healthcare, an eight-hospital integrated health system in Pennsylvania, and Torrance Memorial Medical Center, a 512-bed hospital in California.

Regard will build its newest functionality, named Max, on GPT-4. Max can be thought of like an “AI doctor assistant,” Ben-Joseph said. It is being designed so that clinicians can ask the tool things like “What medications was this patient on this week?” or “Can you give me a summary of what was going on with his patient before they were admitted?” and get instant answers. 

The startup’s new pilot LLM functionality will automatically draft portions of clinical notes by harnessing data from the EHR. It will also search through patient data rapidly to provide clinicians with an autocomplete functionality that helps them save time as they finish up documentation.

Healthcare’s generative AI market is quickly becoming a crowded one, with more and more companies rolling out tools to help alleviate clinical burnout. But Ben-Joseph contends that Regard can stand out from other companies in the space, such as Abridge and DeepScribe.

“We’re quite different from a lot of the scribing companies. They will listen to the conversation and try to transcribe it, but we’re focused primarily on inpatient medicine, where patients are very sick, and there’s a lot of data that is gathered,” he explained. 

Regard is hoping its new partnership with OpenAI will help its technology stand out from the crowd even more. The startup signed a business associate agreement with OpenAI, which allows the parties to exchange data in a HIPAA-compliant manner. The collaboration also gives Regard access to more computing power as it builds out its capabilities.

“The reason we chose OpenAI in particular is because their GPT-4 system is the state of the art right now. We did testing with various other options — we looked at GPT-4, we looked at LLaMA (sold by Meta), we looked at some of the other open source models. GPT-4 gave the best answers, and when you’re working with real patients and doctors and the jobs they do, you need the best of the best,” Ben-Joseph declared. 

OpenAI says that GPT-4 is “10 times more advanced” than its predecessor, GPT-3.5. For instance, GPT-4 passes a simulated bar exam with a score on par with the top 10% of test takers, whereas GPT-3.5’s score was around the bottom 10%. That explains some of the interest from companies like Regard.

“Our mission is to help physicians do the best work that they could possibly do and empower them with technology. This new functionality is something that adds to that mission,” Ben-Joseph declared.

Photo: hirun, Getty Images



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *