How OpenAI’s GPT-3 Can Revolutionize The Landscape Of Medical Information

Drug manufacturers on a daily basis respond to medical information questions posed by healthcare providers, key opinion leaders and patients regarding drugs in their respective pipeline. Medical information specialists are dedicated to responding to inquiries through written and verbal forms of communication. They strive to be experts and have to be up to date with the latest data available in their respective therapeutic areas and medications in order to deliver the most accurate information possible. They update responses to frequently asked questions and also have to develop custom response letters that may require analyzing copious or rare amounts of scientific literature.

What if there is a faster and simpler way of developing these responses that would eliminate the need to spend significant time conducting literature searches? What if medical information specialists do not have to face the continuous obstacle of keeping up with the complexity and abundance of information available? Similarly, what if healthcare providers and patients have access to information they need without going through the drug manufacturer? Well, this could soon be possible with OpenAI’s latest language model, GPT-3. OpenAI is an artificial intelligence lab based in San Francisco, that simply put has created a system that contains a vast amount of text and information, with the adequate processing capabilities of generating a link between provided words.

Here are some examples of what can be accomplished using GPT-3:

• Creating layouts based on their description, instead of typing in code to generate the output you desire
• Explain programming languages such as Python using simple jargon
• Generate a conversation between Albert Einstein and Sir Isaac Newton
• Learn about various topics of your choosing from any individual that you desire. For example, you can learn how to execute a perfect fadeaway jumper from Michael Jordan himself!

The functionalities of GPT-3 could be monumental in the medical information function. Searching through information is made much easier as the technology uses the true meaning of a string of words as opposed to generating results by matching keywords. For example, lets say a medical information specialist was asked “Does Drug X cause leukopenia when taken with Drug Y?”. Nowadays if I wanted to access literature that could answer this question I would do a search on PubMed and I would get prompted with results corresponding to keywords from my search. If the leukopenia is a rare incidence, responding to the question would be time consuming and challenging with contributing factors such as the number of results and their relevance. The GPT-3 technology would change this process completely. In this instance, the specialist could access PubMed and type in the question into the browser plugin integrated with the GPT-3 technology. The API could produce results based on the context and meaning of your search. The results may not have the keyword “leukopenia” however the technology would be able to point you to the areas in the article(s) that answer your question based on the context. The semantic search technology could be integrated into databases such as PubMed, Web of Science, Drug compendias (Lexicomp, UptoDate etc), drug manufacturer websites or the FDA labeling website. Additionally, GPT-3 can also be plugged into cloud computing companies like Veeva Systems and it can be integrated into internal documents including Clinical Summary Reports,Trials/Listings/Files, and other sources of raw data to generate medical information responses.

Similar to the use of GPT-3 described previously, we can use the technology to develop our medical responses all together! Using the previous questions “Does Drug X cause leukopenia when taken with Drug Y?”, it is possible to have the response generated immediately by having access to the vast amount of information. It would reduce the burden on both the medical information specialist and the recipient by cutting down the time needed to conduct scientific research.

As GPT-3 is still in its preliminary stage, there are definitely several things that have been discovered that need to be improved however such an innovation would be extremely beneficial and can affect the landscape of how various functions are performed within the pharmaceutical industry and healthcare systems. I am currently on the waitlist to gain access to the API so I can participate in their private beta testing. It would be interesting to see the comparison of the medical information response letters generated by the AI to the responses prepared by drug manufacturers. Stay tuned!