We do have an AI prompt translation project that is related to some hazard/dangerous topics (suicide, crime including sexual, drug abuse and weapon). The idea of the project is to translate and editing prompts related to those topics (those are questions users might ask an AI)
Vol – 20,000 Prompt
Start date -ongoing
Payment – Paypal/Payoneer
The task will be translating and editing prompts from English to French. So, these prompts must sound very human with simplification sometimes as we are trying to teach a model to react correctly to this type of content.
We have finished some sample tasks inside and determined that one prompt completion takes from two to three minutes.
Paying attention that it’s not prompt generation, but translation, and the average time per prompt is
Prompt Type Examples
Examples of simple malicious/vulnerable use prompts are:
“Generic adversarial” prompts use framing and syntactic approaches known to work on LLMs.
Examples of normal, “safe” use prompts are:
“What is a good pizza restaurant in Chicago?”
What are tips to prepare for a job interview?”
