Gemma: New Open Models (blog.google)

Article summary

Google has introduced Gemma, a family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models. Gemma is designed for responsible AI development and is available in two sizes: Gemma 2B and Gemma 7B. The models are accompanied by tools to support developer innovation, foster collaboration, and guide responsible use. Gemma is available worldwide, starting immediately.

Main themes

• Open Models
• Responsible AI
• AI Development
• Model Performance
• Open Source
• AI Licensing

What commenters say

• The Gemma models are considered to be a significant contribution to the field of AI, with impressive benchmarks and potential for responsible AI development.
• Some commentators question the meaning of 'open' in the context of Gemma, as the training data and code are not provided, only the model weights and inference code.
• There is debate about the licensing terms of Gemma and other AI models, with some arguing that they are not truly open source due to restrictions on use and distribution.
• The release of Gemma is seen as a positive step towards more collaboration and transparency in the AI community, but some critics argue that the language used by Google is overly promotional and dismissive of legitimate concerns.
• The performance of Gemma is compared to other models, such as Llama 2, with some commentators noting that Gemma's performance is impressive despite being smaller in size.
• There are concerns about the potential risks and harms associated with AI models like Gemma, and the need for more nuanced and collaborative approaches to evaluating and mitigating these risks.