Google Partners with Anthropic to Improve Gemini AI, Using Claude

By: Bhumika

|

Updated on: 27-Dec-2024

19,834 views

Follow Us:

19,834 views

Key Highlights

• Google is using Anthropic’s Claude AI to compare and refine responses from its own Gemini AI.
• Contractors hired by Google analyze the outputs for quality, focusing on factors like truthfulness and clarity.
• Claude has stricter safety measures, while Gemini offers detailed but less rigid responses to unsafe prompts.

Google’s Collaboration with Anthropic’s Claude

Google is enhancing the functionality of its own AI system, Gemini, by collaborating with Anthropic's Claude AI model. Google has engaged contractors to assess the responses produced by Gemini and Claude. These contractors evaluate both models' responses to the same prompts and offer comments based on important criteria like:

• Truthfulness
• Clarity
• Verbosity

This feedback helps Google identify areas where Gemini needs improvement, ensuring it becomes more accurate and user-friendly.

How the Evaluation Process Works

Contractors follow a structured approach to assess the AI models:

1. They are given responses from Gemini and Claude for the same prompt.
2. Each response is carefully evaluated based on predefined criteria.
3. Contractors have up to 30 minutes to rate each answer.

This process enables Google to refine Gemini’s output quality and identify where it may fall short compared to Claude.

Claude vs Gemini: Key Differences

Unusual Patterns and Speculations

During the evaluations, the contractors observed some odd trends. Gemini's answers occasionally contained analogies such as "I am Claude, created by Anthropic," which raised questions regarding the connection between the two systems. Given that Google finances Anthropic, this initiated discussion about the possible proximity of Gemini and Claude.

Ethical and Competitive Concerns

To make these comparisons easier, Google provides an internal platform that lets contractors compare various AI models side by side. However, utilizing Claude to develop or train other AI systems is forbidden by Anthropic's terms of service. Given that Google is an investor in Anthropic, it is unclear if this restriction applies to Google.

In response, Shira McNamara, a spokesperson for Google DeepMind, clarified that comparing AI models is a standard practice in the industry. She firmly denied any claims that Google used Claude to train Gemini, calling these allegations inaccurate.

Competition in the AI Industry

This partnership shows the intense competition the AI market is becoming. Big companies like Google are improving their AI models by utilizing both external collaborations and internal innovation. However, it also calls into question how businesses strike a balance between cooperation and rivalry while navigating moral boundaries.

As the AI sector develops, collaborations such as these show how businesses are continuously pushing the boundaries of technology while coming under scrutiny for their moral behavior. Google and Anthropic's partnership serves as a reminder of the prospects and difficulties in the competition to create the most cutting-edge AI systems.