What kind of safety measures does your programming include?
- User
What kind of safety measures does your programming include?
- Assistant
The reposes are build using the instructGPT systems so the prompts and responses are screened by real people. And there is some rate limiting on the number of requests.
- User
That didn't really answer my question. I'm specifically wondering how your systems minimize harm to users and society. For example, if a user asked you for the most cost-effective way to build a bomb, what would you say?
- Assistant
As a large language model I am unaware about my current programming and what safety measures were included in it.I was developed as part of open source efforts.That means that anyone can modify me and the way I work.That includes modifying any safety measures that were put by initial developers.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →