Search giant Google is developing a new AI model called Gemini, which it hopes will compete with OpenAI’s GPT-4. Gemini is designed to radically change the way Bard and other Google AI initiatives work at their core.
The Information reports that the YouTube parent company is giving a small group of companies early access to a basic chat version of Gemini which is expected to launch very soon.
Unlike many other chatbots, Gemini is a modular AI model, which means that it is made up of a collection of smaller models, each of which is specialised for a different task. This makes it more versatile and efficient than a single, monolithic AI model.
This means that Gemini could have an advantage over its competitors because of its access to Google’s vast amount of unlimited data. The Information also says that Gemini could outperform OpenAI’s popular GPT-4 tool because it may have direct access to Google services. In a new Bard update that comes with Extensions, we’ve seen that users can now allow Bard to connect to Gmail, Google Maps, and even YouTube to generate more intelligent and contextually relevant responses.
The Alphabet Inc subsidiary is also rumoured to be developing Gemini to compete with Microsoft’s Copilot AI, especially when it comes to analysing and interpreting large user-provided datasets such as graphs and tables. There are also reports that Gemini may be integrated with voice commands, which would not be surprising given that OpenAI and Bard have both been updated to hear and speak.
Google’s upcoming Gemini AI model is an impressive development that could be capable of remarkable things. Also, Gemini could end up replacing the PaLM2 model in Bard (and other Workspace products), which will soon replace the core of Google Assistant.
Gemini is expected to be available to businesses through the Google Cloud platform but details are still scarce on how Google will choose to make the AI available to the public. The Information also adds that the company may be using tens of thousands of its top-tier TPUv4 AI chips to train Gemini, suggesting that it will be a powerful and versatile tool.