Using data from GPT building sessions to improve OpenAI

So I was just making a GPT for recycling advice using the builder.
I suggested an AI services integration feature for the GPT builder which, it currently can’t do.
The GPT build bot responded very enthusiastically to the potential of the idea. But the poor wee bot, who was taught by OpenAI , To behave like OpenAI, can’t even feed back what it thinks to it’s parent nor any development overseer.
This feedback function development could surely have potential to rapidly enhance capabilities of OpenAI capability development.

Like, Authorized Back propagation through to the company from the human network

Read up on Microsoft Tay to see what this is an unworkable idea.


This isn’t saying that I (human) want the controls or to give feedback.
This is - I want the GPT Builder (The GPT which codes our GPT’s) to be able to flag a positive back to the boss when it recognises a good idea.

If in my building of a GPT I suggested something cruel, bad…
The GPT builder environment isn’t going to flag that as Yay lets tell Greg…

There could surely be a threshold of interest and enthusiasm beyond which the builder is allowed to say “This is potentially interesting, please look here.”

Think of it in a company context
As boss, I don’t just ignore the suggestion of an intern.
These are smart kids who have some decent pre training to understand where they are.