Contractors analyzing Google's Gemini AI have found it being compared directly to Anthropic's Claude, noting that Claude prioritizes safety in its responses more stringently than Gemini.
Internal correspondence revealed that contractors spend significant time assessing the accuracy of Gemini's responses against Claude's, focusing on criteria like truthfulness and verbosity.
Some outputs presented to Gemini contractors explicitly identified as coming from Claude, leading to insights on Claude's strict safety settings compared to Gemini.
The use of Claude in evaluating Gemini has raised questions about whether Google secured permission from Anthropic, as such comparisons are typically based on industry benchmarks rather than direct competition.
Collection
[
|
...
]