My understanding is that the OpenAI Microscope (is this what you meant by microscope AI?) is mostly feature visualization techniques + human curation by looking at the visualized samples. Do you have thoughts on how to modify this for the text domain?
Microscope AI as a term refers to the proposal detailed here, though I agree that I don’t really understand the usage in this post and I suspect the authors probably did mean OpenAI Microscope.
We meant the linked proposal. Although I don’t think we need to do more than verify a GPT’s safety, this approach could be used to understand AI enough to design a safe one ourselves, so long as enforcing modularity does not compromise capability.
Interesting stuff!
My understanding is that the OpenAI Microscope (is this what you meant by microscope AI?) is mostly feature visualization techniques + human curation by looking at the visualized samples. Do you have thoughts on how to modify this for the text domain?
Microscope AI as a term refers to the proposal detailed here, though I agree that I don’t really understand the usage in this post and I suspect the authors probably did mean OpenAI Microscope.
We meant the linked proposal. Although I don’t think we need to do more than verify a GPT’s safety, this approach could be used to understand AI enough to design a safe one ourselves, so long as enforcing modularity does not compromise capability.