23/2/2024
February 23, 2024
Google's new open source AI models, Gemma 2B and Gemma 7B, aim to outperform competitors like Meta and Mistral AI. Designed for researchers and developers, these models are part of Google's strategy to accelerate AI innovation and emphasize responsible development.
In a strategic move that underscores its commitment to the rapidly evolving AI landscape, Google has announced the launch of two open source AI models named Gemma 2B and Gemma 7B. This development is seen as a direct challenge to industry giants like Meta and the French open-source model pioneer, Mistral AI. Gemma, derived from Google's earlier Gemini project, signifies a major leap in the company's AI offerings, aimed at bolstering its position in the competitive AI market.
Google's introduction of Gemma comes at a time when the tech giant is keen on making up for lost ground in the AI race. The two models, Gemma 2B and Gemma 7B, are built on the foundational research and technologies that powered the Gemini models, indicating a continuity and evolution in Google's AI development strategies. Available with a commercial license, these models are specifically designed for researchers and developers, enabling them to craft their own AI-driven applications.
Gemma will be accessible through various platforms including Kaggle, Hugging Face, Nvidia's NeMo, and Google's Vertex AI, ensuring widespread availability for the AI community. In addition, Google has pledged up to $500,000 in Google Cloud credits to support academic research, further demonstrating its commitment to accelerating innovation in the AI field.
Google's Gemma sets its sights on outperforming competitors like Meta's open source Llama 2 and Mistral AI in various benchmarks, including the Massive Multitask Language Understanding (MMLU) test. Gemma 7B, in particular, showcases superior performance in tasks such as written comprehension and Python code generation, positioning Google as a formidable player in the open source AI arena.
Google emphasizes the importance of responsible AI development, highlighting the careful selection and adaptation of training data to ensure safety. The company has also released a Generative AI Toolkit to guide developers towards ethical use of open models like Gemma. This toolkit provides resources on security standards, safety parameter tuning, and learning interpretation tools to address potential issues, underscoring Google's holistic approach to AI responsibility.