TLDR:
- Google launches Gemma, consisting of new open AI models.
- Gemma is built for responsible AI development using technology from Gemini models.
Tech giant Google, in partnership with Google DeepMind, has unveiled Gemma, a new venture into open AI models. Gemma consists of model weights in two sizes, Gemma 2B and Gemma 7B, released with pre-trained and instruction-tuned variants. These models are designed for small work-based tasks like chatbots or summarizations and can run on various platforms such as laptops, workstations, and Google Cloud. Google has optimized Gemma models using NVIDIA GPUs and Google Cloud TPUs for industry-leading performance.
Gemma is part of Google’s commitment to making AI helpful for everyone, designed to assist developers and researchers in building AI responsibly. Built from the research and technology used for Gemini models, Gemma can be fine-tuned for various business use cases and run on different platforms. The AI is being rolled out on a wider scale and is optimized across hardware, partnering with NVIDIA for performance and usability on various devices.
In addition to the Gemma launch, Google has released a new Responsible Generative AI Toolkit to help developers prioritize building safe and responsible AI applications. While Gemma models are open, the AI itself is not open source. Google aims to support open models for AI development, enabling developers to harness technology in a safe and responsible manner.
CEO of Google DeepMind, Demis Hassabis, expressed pride in releasing Gemma, highlighting the lightweight open models as best-in-class for their size. By incorporating the same technology as Gemini models, Gemma aims to drive rapid research progress in the field of AI development.