OpenAI Revealed to Have Had ChatGPT Chats Reviewed Under Project Lily
A newly published report reveals that OpenAI assigned hundreds of contractors to review real user conversations in order to improve artificial intelligence responses.
Recent reports show that OpenAI has been running a program called Project Lily to improve the response quality of its artificial intelligence model, with hundreds of contractors manually reviewing real ChatGPT conversation logs as part of the initiative.
Project Lily and Human Review
Details of the human review process conducted by OpenAI to improve model responses have come to light.
It was understood that within the framework of an operation named Project Lily, real-world conversations were examined by humans.
Hundreds of Contractors on Duty
According to published reports, the company hired hundreds of contractors to review chat logs.
These employees evaluate anonymized real user chats to measure the system's performance.
Review Criteria and Evaluation
Operators, referred to as prompt reviewers, judge the response quality of the artificial intelligence based on specific parameters.
The texts are checked for the presence of excessive artificial intelligence language, sarcastic tones, emojis, and contrived manners.
Personal Information and Privacy Concerns
It was reported that during these manual review processes, some personal information shared by users could also be viewed by human reviewers.