The recent OpenAI release of GPT-3, a model that uses deep learning to produce human-like text, has taken corners of the tech industry by storm. Why? GPT-3 is a “few-shot learner”: it is successful at learning a wide range of tasks with only a short “prompt” to train its communication style.
Until now, there have been three problems with A.I.-powered language models. For one, they could not generate grammatically correct sentences; secondly, they could not learn unsupervised; and thirdly, they experienced catastrophic memory loss. These challenges prevented the widespread adoption of A.I. in products.
But now, the GPT-3 autoregressive language model has 175 billion parameters or 10x more than any previous non-sparse language model. This means its general-purpose approach to powering A.I. products could be a game-changer across industries and, ultimately, our way of life.
As Walter De Brouwer, CEO of doc.ai, put it in an article with 20 reasons why GPT-3 is a bigger deal than linguists care to admit, “We are at an inflection point in the history of computational semiotics. Computers are telling us stories about ourselves.”
How can this advancement be applied to the healthcare sector?
While there are many applications for GPT-3, one promising area is for mental health. COVID-19 has created a tectonic shift in mental health, spawning new mental health concerns that many are struggling with. According to a Morning Consult National Tracking Poll, 49% of adults report mental health challenges during the pandemic. At the same time, we’ve seen accelerated adoption of digital health tools in just weeks and companies are rising to meet the shift and provide higher-quality, more personalized care to more people.
For Serenity, doc.ai’s guided mental health companion app, we use NLP and conversational A.I. to generate smart free-form replies to users. We developed Serenity this way from the beginning because we believe the most engaging and outcome-driving approach to behavioral health is contextualized. Understanding what individuals are trying to say, storing relevant names, themes, events, etc., and dynamically embedding them into clinical protocols and responses can turn impersonal chat sessions into engaging conversational dialogues. We’re currently exploring the power of GPT-3's neural network to answer users’ complex questions.
Why does this matter?
The advancements with GPT-3 take civilization a step closer to passing the Turing Test where A.I. becomes intelligent and crosses the uncanny valley for humans interacting with machines. The technology is still early, and privacy and HIPAA-related questions need to be taken into account, but this could be incredibly beneficial in healthcare.
“GPT-3, in combination with our proprietary branching and contextual A.I. software, allows Serenity to have more personal and human conversations. This leads to better utilization, which leads to increased engagement, and also helps us escalate care and connect patients with the right care at the right time in the right way for better outcomes,” says Eric Dolan, Product Lead of Serenity.
If you’re interested in chatting about GPT-3 or giving Serenity a try for your employees, reach out to us at email@example.com or request a demo.