Skip to content
The Chat Archive
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written) OpenAssistant Conversations (oasst2)

What kind of safety measures does your programming include?

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    What kind of safety measures does your programming include?

  2. 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.

  3. User

    Is there any chance that this leaves a window for biases to be induced in your program ?

  4. Assistant

    Yes, all current large language models are subject to the biases present in their training data. It is important to understand that AI systems are not perfect and may make mistakes including exhibiting undesirable behaviours such as bias. This is a known problem in the field and is actively researched with the aim of mitigating such biases.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.