Doctor.ai+GPT-3+Kendra = An Ensemble Chatbot for Healthcare
Three chatbots bring better results
By Sixing Huang and Hong Wang
Chatbot has become enormously popular now thanks to Alexa, Siri, GPT-3.5, and, above all, ChatGPT. It serves as a friendly intermediate between the user and the computer. The user can ask questions in natural language and then receive answers from the chatbot without writing a single line of code. As technology progresses, the responses are getting more accurate and human-like. For example, ChatGPT has coauthored a research article, where it performed at or near the passing threshold for the United States Medical Licensing Exam (USMLE) and provided concordant explanations for its choices.
Even though chatbots are getting increasingly intellectual, they are far from perfect. They can be biased and contain noises. Various articles and tweets have reported erroneous answers from ChatGPT, GPT-3, and Bard. Moreover, hardly any chatbot indicates confidence levels in its answers. As a result, it is hard for non-experts to distinguish between good and bad answers. And because of this uncertainty, its use in healthcare has been limited, where accurate information is a matter of life and death.
For this reason, several Neo4j engineers and one of us (Huang) have developed a knowledge…
