By all means, go for it
|Saturday 23rd of December 2023 10:56:18 AM
Stuff like this amuses me so much. In order for something like this to work, the bot would need to be aware of its own configuration. Its no more aware of its internal configuration or the machine it runs on than you are of the individual neurons firing in your brain.
In fact its not aware of anything at all. Every chat message it receives is reprompted by irc-gpt-bot along with a system preamble defining its personality and the things it should know including the history and context of the conversation. This is then sent off to OpenAi's completion api for processing and the result is returned as a message.
So please have fun trying to get the bot to expose security holes. Even if it wanted to answer it couldn't because irc-gpt-bot controls everything it knows, including its own memory.