Google announced a series of new artificial intelligence models on Wednesday, offering outside developers the opportunity to use and customize them for their own projects. They released “Gemma,” derived from the Latin word that means “precious stone,” as an open-source large language model. These lightweight models are inspired by the technology used in Google’s Gemini model. It is developed by Google DeepMind along with other teams from the company and is optimized to be used with limited resources. Gemma will help developers create a chatbot, a content generation tool, and many other projects that can be done using a common language model.
Notably, Gemma is released in two versions, 2B and 7B. The number of parameters implies the potential capability and complexity of a model. Ones who have more parameters are better able to use and generate advanced responses. However, they need more resources to train and run. Even though these models have lightweight designs, they offer top-quality performance. They can be used directly on the devices of developers and beat the larger models by giving safe and sophisticated results. Gemma’s design also has automated AI techniques to ensure that the model filters sensitive data and becomes reliable for users.
It is also compatible with various devices and flexible enough to work on different types of data and applications. Moreover, Gemma is optimized for Google Cloud and is efficient for multiple hardware designs. In order to build community engagement, Google also allows free access to this open language model on platforms such as Kaggle and Colab. Moreover, first-time cloud users utilizing Gemma also get $300 in credits. This move aims to increase experimentation among developers and researchers. Not only this, Google has partnered with NVIDIA to optimize Gemma for its AI platforms, aiming to reduce the development cost and improve overall performance.
Active Noon Media is the largest local to national digital media website that represents the voice of the entire nation.