Google is reportedly comparing Gemini’s answers with those produced by the human Claude artificial intelligence (AI) model. According to the report, the Mountain View-based tech giant asked third-party contractors to evaluate the quality of Gemini’s response to assess whether Cloud’s response was better. While this is not a rare practice, using third-party AI models for this purpose requires permission from the company. The report cannot confirm whether the tech giant has obtained any permission from humans.
Gemini’s use of Claude is reportedly improved
According to a report by TechCrunch, contractors engaged in Gemini compared their output to contractors created by Claude, the former’s direct rival. The publication claims to have seen internal letters between Google and the contractors, who were told to follow this practice.
The report also claims that contractors allow up to 30 minutes per prompt to determine whether they prefer Gemini or Cloud’s response. It is worth noting that these contractors are often subject matter experts who evaluate chatbots’ answers to niche and technical topics such as authenticity, precision and lengthy such as several parameters such as authenticity, precision and lengthy.
After the practice began, some contractors reportedly began to notice references to humans and Claude in Gemini’s answers. The publication claims that a Gemini output directly says: “I am Claude, created by humans.” Usually, this should not happen if only Gemini’s answer is compared with Claude’s answer. This has attracted the attention of whether developers feed Cloud’s production to Gemini when the former responds better.
As far as its business conditions are concerned, humans stipulate that those who visit Claude are prohibited from establishing “competing products or services” or training “competing AI models” without human approval.
While the department did compare the model outputs for evaluation purposes, Gemini was not trained in anthropomorphic AI models, speaking with TechCrunch.
Last week, a report claimed that contractors assigned to evaluate Gemini’s responses were told to rate the output of chatbots, even if they were not professional. While earlier, contractors could skip such questions, it was reportedly selected.
