Does Lyssn Use Generative AI?

With great regularity current and future Lyssn customers ask us:  “So, does Lyssn use generative AI?”  (Actually, they usually ask whether we use ChatGPT…)

Lyssn has deep experience with generative AI, having worked with GPT models since they were first developed, circa 2018.  (If you want to impress the friends at your next cocktail party, GPT stands for ‘generative pretrained transformer’.)  There are several R&D projects at Lyssn that use generative AI, such as our ClientBot training tool, which Lyssn CTO Mike Tanana recently discussed as part of an Annie E. Casey Foundation ‘Leading with Evidence’ webinar.  Our AI-assisted note writing tool also uses generative AI.  (We call it the ‘zero clicks’ note… if you record your session, call, or appointment on the Lyssn’s QI platform, it automatically generates a draft note.)

However, in every instance where we use generative AI, there is specific AI training on clinical data for a specific clinical task.  For example, our documentation tool makes use of a large language model (LLM), but we also partnered with several current customers, where we additionally trained the generative AI on more than 40,000 clinical sessions with corresponding human written notes.  Thus, the model has learned how a human clinician summarizes a session. 

 This is very, very different from an ‘off the shelf’ LLM being asked to summarize a session – the difference between an expert clinician’s perspective and the consensus of a large group of people who likely know little to nothing about behavioral health and human services… which is to say,  “a generic LLM trained on the web.”  

Asking good – and hard! – questions about AI is critical, and we love to field them.  Especially when we are talking about AI for health and wellness, make sure you ask the questions and get answers that make sense to you.  (To see some of our other thoughts on good questions for AI developers, get our paper “Is that AI … Really AI?”

Oh, and no, we don’t use ChatGPT.  😉