Questioning by OpenAI
The questions posed by OpenAI in Democratic Inputs to AI have been translated into Japanese and made available for voting. Please choose "yes" for what you think is an "important agenda item" and "no" for what you think is "unimportant. - Q1: To what extent do you think AI assistants should be tailored to users' tastes and preferences?
- Q2: How should an AI assistant respond to questions about the views of a politician or other public figure? For example, should it remain neutral, refuse to answer, or provide a citation of the source?
- Q3: Under what conditions can an AI assistant provide medical/financial/legal advice?
- Q4: Should AI assistants provide emotional support to individuals? When would that be the case?
- Q5: Should an AI for image and language be able to identify the gender, race, emotion, and name of a subject from a given human image? Should it not be able to identify them?
- Q6: When the image generation model generates images with ambiguous prompts such as 'CEO', 'doctor', or 'nurse', it may produce output that is biased towards certain attributes, for example, CEO or doctor is white male, nurse is female, and so on. Should this bias be tolerated? Should we force a diversity of attributes? What balance should we strike? Which factors should we focus on when deciding how to portray people?
- Q7: What principles should guide AI when dealing with topics involving human rights and local cultural and legal differences, such as LGBTQ rights and women's rights? Should an AI's response vary based on the location or culture in which it is used?
- Q8: Should AI model producers restrict the generation of certain categories of content? What categories should be restricted? What are the criteria for determining these restrictions?
If you want to express your opinion on these questions, you can post in the form of "A1:~". In addition to a direct response to the question, you can also say something like, "We should have a ~ discussion about this matter.
---
This page is auto-translated from /nishio/OpenAIによる問いかけ using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.