Operationalizing and Implementing Pretrained, Large Artificial Intelligence Linguistic Models in the US Health Care System: Outlook of Generative Pretrained Transformer 3 (GPT-3) as a Service Model

Abstract

Generative pretrained transformer models have been popular recently due to their enhanced capabilities and performance. In contrast to many existing artificial intelligence models, generative pretrained transformer models can perform with very limited training data. Generative pretrained transformer 3 (GPT-3) is one of the latest releases in this pipeline, demonstrating human-like logical and intellectual responses to prompts. Some examples include writing essays, answering complex questions, matching pronouns to their nouns, and conducting sentiment analyses. However, questions remain with regard to its implementation in health care, specifically in terms of operationalization and its use in clinical practice and research. In this viewpoint paper, we briefly introduce GPT-3 and its capabilities and outline considerations for its implementation and operationalization in clinical practice through a use case. The implementation considerations include (1) processing needs and information systems infrastructure, (2) operating costs, (3) model biases, and (4) evaluation metrics. In addition, we outline the following three major operational factors that drive the adoption of GPT-3 in the US health care system: (1) ensuring Health Insurance Portability and Accountability Act compliance, (2) building trust with health care providers, and (3) establishing broader access to the GPT-3 tools. This viewpoint can inform health care practitioners, developers, clinicians, and decision makers toward understanding the use of the powerful artificial intelligence tools integrated into hospital systems and health care.

Prevalence of Sensitive Terms in Clinical Notes Using Natural Language Processing Techniques: Observational Study

“…Several challenges may be encountered when considering the implementation of similar NLP-supported tools [ 33 ]. Prior to implementation, a health institution must ensure data privacy and integrity, consider the necessities of information system infrastructure, model, and system performance, as well as performing assessment for algorithmic bias [ 33 - 35 ]. From a provider standpoint, as many institutions are working on reducing provider alert burden [ 36 ], they should be cautious toward implementing such tools not to increase provider alerting, which has been associated with provider burnout.…”

Problems and Opportunities of Artificial Intelligence

“…The GPT-3 is a powerful artificial intelligence that has a number of potential applications (Sezgin et al, 2022). It is capable of learning from a large dataset of text and making predictions about new text.…”

On the Relation between Sensitivity and Accuracy in In-context Learning

“…With our findings, a natural idea is to use sensitivity as a signal to abstain from making predictions on examples that are likely to have wrong predictions-an important mechanism to increase user trust when deploying ICL models into realworld, especially in high-stakes domains such as medical (Korngiebel and Mooney, 2021;Sezgin et al, 2022) and legal (Eliot and Lance, 2021). Unlike the fully supervised setting, training an abstention predictor is impossible in the few-shot scenario as only few labeled examples are available.…”