To catch up with ChatGPT, Google reportedly cribbed OpenAI's homework to boost Gemini's AI game

Google has often been viewed as the company holding the shorter end of the stick in the generative AI race despite its vast cloud computing and talent. Microsoft CEO Satya Nadella claimed that the software giant had already missed its opportunity in the AI space, prompting Alphabet's CEO Sundar Pichai to fire back with lethal jabs aimed at the tech giant's fraying partnership with OpenAI:

"I would love to do a side-by-side comparison of Microsoft's own models and our models any day, any time. They're using someone else's models."

But as it now turns out, it might have been a case of the pot calling the kettle black. According to documents exclusively reviewed by Business Insider, Google's contractors at Scale AI reportedly used ChatGPT to train and improve Bard (now Google Gemini).

The damning report further revealed that the contractors generate thousands of responses from ChatGPT, comparing them with to Bard's output. They used the responses to improve Bard's responses, making them at least on par with ChatGPT.

Admittedly, Satya Nadella claimed OpenAI has a 2-year runway to build and develop ChatGPT uncontested, potentially making it difficult for its rivals like Copilot and Gemini to play catch-up.

Scale AI managers acknowledged that OpenAI's chatbot generated more effective responses with improved formatting and interesting facts. The contractors were tasked to make Bard better than GPT-4 with a 15% bonus dangling over their heads for responses that blow ChatGPT out of the water.

It's worth noting that OpenAI's terms of service blatantly prohibit the use of its output to train rival models. However, Scale AI has dismissed the claims:

"Scale did not, and does not, use ChatGPT responses to train Gemini or any models."

Scale AI indicated that the documents obtained by the outlet represent "standard side-by-side evaluations," which are often misconstrued as the actual use of ChatGPT for model training and development. The AI firm further claimed the evaluations are part of industry standards.

Meta's new $14.8 billion stake in Scale AI drives rift with Google

Recently, details emerged about Meta's plan to partially acquire Scale AI in $14.3B deal, giving it 49% ownership of the AI startup. The deal pushed the AI startup's market valuation to $29 billion. It's also worth noting that the Facebook maker also hired Scale AI founder Alexandr Wang to lead its new superintelligence unit.

However, it now appears that Meta's newfound partnership with Scale AI doesn't sit well with Google. The software giant is reportedly set to sever its ties with Scale AI (via Reuters). Google is Scale AI's largest customer, which planned to pay up to $200 million to Scale AI in 2025 for its human-labeled training data, serving as an important element in the development of AI models that power Gemini.