Need Sources Cited by Browsing Plugin Listed Beneath Answer

I just got a really good description of a high profile government official from GPT 4 with the browsing plugin. Unfortunately, it cites its sources using links formatted like footnotes but without any footnotes at the bottom. I need the results formatted with the URLs listed at the bottom unless it would be possible to make an API call asking for HTML output so that when inserted into a database it will echo out on a web page just fine.

There’s currently no way of using the browsing function or other plugins though the API

You can however click the number you mentioned to go directly to the website and copy the url yourself :hugs:

The whole point of using the API is to develop automated solutions which don’t require you to copy and paste anything yourself. In my case, I have a list of government officials who I ask it to research. The goal is to compile links to relevant online articles about them for the purpose of improving transparency and accountability. People shouldn’t have to do so much research just to know if an official has ever been publicly accused of misconduct.

The list has several thousand names, so I built an app to iterate through them. Unfortunately, 3.5 prefers to make up fake stories with fake sources, so it is a no go. I need something just like the API but with responses from the browsing plugin.

I am specifically not looking for personal information (ex: address, phone number), but 3.5 almost always responds with a refusal to research people in general. GPT 4 with browsing at least searches its memory and online sources before compiling an accurate summary about that official.

I can understand what you’re trying to do:

I’m also a big fan of transparency and accountability. But I’m also a fan of having a human in the loop when dealing with sensitive subject’s

It sounds like you want a fully automated solution.

Utilizing AI to extensively research individuals raises ethical considerations including potential privacy infringement and doing so completely automatically will lead to accuracy issues, this could potentially lead to misinformation propagation, and the risk of creating an unfair, biased depiction due to the impersonal and uncontextual approach.

Again, this is not possible at the moment and there’s no timeline for the release of such functionality.