Google on Wednesday introduced a set of new artificial intelligence models named Gemma, which developers outside of Google can potentially utilize to create their own applications.
The technology giant, a subsidiary of Alphabet, is allowing individuals and businesses to construct AI software based on Gemma, a family of “open models,” without charge. Google is making crucial technical data, including model weights, publicly accessible to support developers in their endeavors.
While the move aims to attract software engineers to leverage Google’s technology, it also serves to boost the usage of its increasingly profitable cloud division. The Gemma models are optimized for Google Cloud, and first-time cloud users receive $300 in credits when utilizing these models, according to the company.
While not fully “open source,” as Google retains some control over terms of use and ownership, the release of Gemma is a step toward increased transparency and collaboration in the AI community.
The decision not to make the larger Gemini models fully open contrasts with the Gemma release. The Gemma models come in two sizes, with two billion or seven billion parameters, representing the values an algorithm considers to generate output.
In a collaborative effort, chipmaker Nvidia announced on the same day that it has worked with Google to ensure smooth operation of Gemma models on its chips. Nvidia also disclosed its plans to make chatbot software, currently in development for running AI models on Windows PCs, compatible with Gemma in the near future.
