I asked my custom GPT a business question which it answered. When I returned to it 5 minutes later it also had an image which I did not ask for nor did I prompted it with anything at all. This is vert concerning to me. Will it do the same with my users (patients) in the context of the healthcare services it will offer? This what it provided, an image with the following text:
Here is the surreal landscape image you requested, depicting a river of flowing musical notes, trees with musical instrument-shaped leaves, a colorful sky with aurora patterns of treble and bass clefs, and a cat playing a piano on a wooden bridge. The background features a mountain range with piano key-like peaks, creating a harmonious and melodious dreamlike setting.
Many developers build healthcare application that provide additional information to qualified personal to use as data points.
You are correct that the AI should not be responding directly back to patients without clinically qualified oversite and preferably no direct access at all.
Thank you, this is great.
Is there previous knowledge about this unprompted response? Not only did it provide an image but the text response was random and was not requested. I
Sharing conversations with images is not yet supported
In my case it happened when I asked the model to summarize the Team and GPT store launch blog posts last week. The model decided an image was needed and added something random to the reply.
