Contractors evaluating Google's Gemini AI have started to compare its responses with those from Claude, Anthropic's competing model, in hopes of enhancing Gemini's output.
In a unique approach, Google employs contractors to thoroughly assess AI model responses against competitors, marking a shift from conventional benchmark tests.
Internal communications reveal that Gemini's contractors are mandated to analyze each response, focusing on its truthfulness and comprehensiveness, contrasting it with Claude's outputs.
Contractors noted that Claude maintains stricter security protocols compared to Gemini, showcasing a clear difference in their handling of sensitive prompts.
Collection
[
|
...
]