One reason I’m critical of the Anthropic RSP is that it does not make it clear under what conditions it would actually pause, or for how long, or under what safeguards it would determine it’s OK to keep going.
Can you link an example of what you believe to be a well-worded RSP?
Can you link an example of what you believe to be a well-worded RSP?
You’re aware that there’s only one public RSP?
You can find the current closest thing various companies have at https://www.aisafetysummit.gov.uk/policy-updates/#company-policies
I never said it had to be an officially sanctioned one, plenty of folks are capable of writing drafts, ideations, conjectures, etc...
If literally no one has ever published something along these lines, then that’s probably the most promising avenue of investigation.