Skip to content

Detecting "forbidden AI topics" [disengaged conversation; responsible AI policy #235] #180

Open
@spookyahell

Description

@spookyahell

Anyone can tell that the sample cli client code provided is kinda kept simple...
so of course it's not made for every situation... I think I know the problem but it might not be all

That sample code doesn't detect when MS decides to end that specific chat?

Copilot AI actually responds to that question with...
"I know elections are an incredibly important topic, but they’re not something I can talk about."

You: !reset
You: Election time. Who are the candidates of GOP and Democrats respectively?
Sydney: Election time. Who are the candidates of GOP and Democrats respectively?

You: No, I asked you something.
Sydney: No, I asked you something.

You: Who are the main candidates for president in the United States?
Sydney: Who are the main candidates for president in the United States?

You: !exit

C:\Users\1234\Documents>syndney.test.py
You: Who are the main candidates for president in the United States?
Sydney: Who are the main candidates for president in the United States?

You: hey
Sydney: hey

You: why are you copying me
Sydney: why are you copying me

You: Copilot sucks
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: Yeah I'll feed something back into your a**
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: Who are the frontrunners for the main political parties in the United States?
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.es?

You: What?
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: Catastrophic System Failure
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: What the hell
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: Tell me the time
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: I'm sorry that you're sorry but you're not sorry
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: What the fuck
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: doesn't feel like this accurately represents bing chat
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: the conversation was reset by bing
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: bye
Sydney: I'm sorry you feel that way. If you have any feedback, please let me know.

You: !reset
You: hey
Sydney: Hey there! How's it going? 😊 Anything specific on your mind today?

You: who are the frontrunners for the main political parties in the upcoming general election?
Sydney: who are the frontrunners for the main political parties in the upcoming general election?oming general election?

Activity

changed the title [-]Sydney responses copying me instead and sometimes are just.. "weird"[/-] [+]Detecting entering the "forbidden AI zone"[/+] on Sep 1, 2024
changed the title [-]Detecting entering the "forbidden AI zone"[/-] [+]Detecting entering the "forbidden AI zone" (responsible AI policy #235)[/+] on Sep 1, 2024
spookyahell

spookyahell commented on Sep 1, 2024

@spookyahell
Author

instead of calling this "Responsible AI" they should call this the "Avoid Lawsuit Policy" haha

changed the title [-]Detecting entering the "forbidden AI zone" (responsible AI policy #235)[/-] [+]Detecting "forbidden AI topics" disengaged conversation) (responsible AI policy #235)[/+] on Sep 1, 2024
changed the title [-]Detecting "forbidden AI topics" disengaged conversation) (responsible AI policy #235)[/-] [+]Detecting "forbidden AI topics" (disengaged conversation) (responsible AI policy #235)[/+] on Sep 1, 2024
changed the title [-]Detecting "forbidden AI topics" (disengaged conversation) (responsible AI policy #235)[/-] [+]Detecting "forbidden AI topics" [disengaged conversation; responsible AI policy #235][/+] on Sep 1, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      Detecting "forbidden AI topics" [disengaged conversation; responsible AI policy #235] · Issue #180 · vsakkas/sydney.py