In the wild world of AI, Google’s Gemini is taking a page from the competition—specifically Anthropic’s Claude. Contractors working on improving Gemini have been comparing its performance to Claude, Anthropic’s own AI model, according to internal emails seen by TechCrunch. While Google’s not spilling the beans on whether they’ve got permission for this comparison, it’s clear the AI battle is heating up.

Typically, companies test their AI models against industry benchmarks. But Google’s contractors are diving deeper, scoring responses based on accuracy, truthfulness, and even verbosity. They’re giving each answer up to 30 minutes to judge which is the superior model—Gemini or Claude. Sometimes, they’ve noticed Claude being extra “safe,” refusing to answer prompts deemed risky (like nudity or role-playing), while Gemini sometimes goes a bit off-script.

There’s a catch, though: Anthropic’s terms of service prohibit using Claude to build or train competing products without approval. Google, a big investor in Anthropic, hasn’t confirmed whether it’s gotten that approval. However, Google insists it’s just using Claude as a benchmark, not for training Gemini.

In the end, it seems like a case of friendly rivalry—at least until the lawyers weigh in.

SEO Keywords: Google Gemini, Anthropic Claude, AI models, AI comparison, Google AI, Claude safety features.

Source Info: https://techcrunch.com/2024/12/24/google-is-using-anthropics-claude-to-improve-its-gemini-ai/

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

Sign In

Register

Reset Password

Please enter your username or email address, you will receive a link to create a new password via email.