Google and Deepmind are working together to develop a GPT-4 competitor called "Gemini".
The Information reports that the Google Brain AI team is working with Deepmind on an alternative to OpenAI's GPT-4. Work on the Gemini project is said to have begun in recent weeks after Google's Bard failed to keep up with ChatGPT.
Together with Deepmind, Google Brain is said to be developing a large language model that will have a trillion parameters like GPT-4. The project is said to be using tens of thousands of Google's TPU AI chips for training. However, it could take months to complete. Whether Gemini will be multimodal is unknown.
Deepmind has also developed the web-enabled chatbot Sparrow, which is optimized for security similar to ChatGPT.
Deepmind and Google Brain had "little choice but to work together"
The London-based AI company otherwise operates almost independently of Google, but employees at both companies agree that "OpenAI has outflanked them", writes The Information. The startup has also poached important researchers at Google and Deepmind.
According to The Information, Google Brain and Deepmind each wanted to develop their own GPT-4 competitors, which would have required enormous amounts of computing power. So, according to a person familiar with the situation, the two had "little choice but to work together."
Even highest-ranking AI researcher at Google writes code for Gemini
The importance of the project to Google is demonstrated by the fact that even Jeff Dean, the head of Google Brain and Google's most senior AI research executive, is taking a technical role in the project and is writing code, the report says.
Away from Gemini, Google is continuing to push the development of Bard and is expected to merge it with Google Assistant, according to a report. The company also plans to bring generative AI to GMail, Docs, Slides, and other products and is making its own AI models available to developers via the cloud.