Google accused of using novices to fact-check Gemini's AI answers

Google criticized for novice fact-checking methods concerning Gemini 2.0's AI answers.

: Google instructed GlobalLogic contractors to not skip prompts lacking their expertise when evaluating AI. Previously, they could skip tasks outside their domain. This change is reportedly to increase evaluation accuracy but has raised concerns. Google has not provided any comments on the issue.

Google has faced criticism for allegedly instructing GlobalLogic contractors not to skip AI evaluation prompts out of their expertise. This directive involves assessing Gemini 2.0, Google's AI project, and the policy shift has prompted debate over the accuracy of such evaluations.

Previously, contractors were allowed to skip tasks requiring critical expertise they did not possess. They were advised to only rate tasks they understood, especially when domain-specific knowledge was essential, sparking confusion amongst evaluators.

This alteration aims to streamline accuracy, yet some contractors argue it could lower the overall quality. Google's silence on the matter leaves the impact of these changes open for interpretation by industry observers.

GoogleAI evaluationGemini