Process for determining accuracy of responses?

Non-technical person asking the question so appreciate the feedback as I’m learning. I typed a few simple questions (related to blood pressure and BMI) that I could imagine patients asking. Responses were good but got me thinking. What’s the process to verify accuracy of response? Is it just a thumbs up or down? Is the persons credentials taken into consideration when the thumbs up/down are aggregated? I saw the info below in FAQ but for health and medical related questions there would need to be a more rigorous vetting of accuracy of the responses. Is there a team working on this? Seems like there’s potential with this as component of our system to address access and costs for patients. If anyone had resources they can share to learn more that would be appreciated. Thanks. Hank

Can I trust that the AI is telling me the truth?

  1. ChatGPT is not connected to the internet, and it can occasionally produce incorrect answers. It has limited knowledge of world and events after 2021 and may also occasionally produce harmful instructions or biased content.We’d recommend checking whether responses from the model are accurate or not. If you find an answer is incorrect, please provide that feedback by using the “Thumbs Down” button.

“not connected to the internet” is different from “my training models use data from Wikipedia and internet resources…etc” It went on to explain that when u query, it is not connecting to internet in real time in order to provide u responses. However, it is a collection of web data, books, texts, documents, periodicals and so on, up thru September 2021. On the other hand, anyone who currently interacts with Chathgpt is broadening its language and training model using your and its inputs. Hope I explained that correctly

I’d say it’s the same process for any resource. Be critical, and confirm the information with multiple reputable sources.

Keep in mind that if you were intending to use it for medical information, you would need to fine-tune your own model using a very reputable source. cGPT is not in any way purposed for medical advice.

1 Like

they just released a plugin feature. Maybe you can introduce your own plugin with your data set. Good luck