Google launches Gemini, its most-advanced AI model yet, as it races to compete with ChatGPT

Google has launched a new AI model, Gemini, aiming to compete in the generative artificial intelligence space and enhance various applications from Google’s consumer apps to Android smartphones. Gemini, which Google touts as its “largest and most capable AI model,” is envisioned to be used across the board from large companies to consumer devices such as the Google Pixel 8 Pro.

Gemini is a multimodal AI model capable of processing multiple types of media including text, images, audio, video, and programming code. The company’s CEO shared his vision of this technological leap representing one of the most significant scientific and engineering endeavors for the company.

A version of Gemini has been incorporated into Google’s proprietary AI chatbot, Bard, with further integration planned for Google’s search engine and Chrome web browser. The introduction of Gemini is part of Google’s broader strategy to bolster its presence in generative AI, following the massive popularity of ChatGPT released by a competitor.

The Gemini 1.0 model comes in three different sizes for various applications: Nano, optimized for mobile devices and app developers; Pro, the default model for an array of tasks and customers; and Ultra, which is described as the most complex AI model Google has created and is still under safety testing.

During the launch, Google highlighted its recent advances in cloud computing, revealing that Gemini was trained using a newer generation of cloud-based processors that significantly speed up AI model training. This development could help the wider AI industry by making AI training more accessible and could potentially strengthen Google’s position in the public cloud services market.

Google claims that its Gemini model has outperformed rival AI models in over two dozen benchmarks that measure an algorithm’s capabilities in areas such as reading comprehension, math, and multistep reasoning.

Despite the advancements, Google is aware of the potential risks associated with AI, such as the possibility of models producing misleading results. The company mentioned that it has taken steps to enhance Gemini’s factuality, though it acknowledges that large language models can still generate incorrect information confidently.

Recognizing these potential risks, Google plans to release Gemini Ultra gradually, first to a select group of customers, developers, partners, and experts for early testing and feedback. This advanced version of the model is also undergoing third-party safety evaluations before its broader roll-out is planned for the following year.