Google has introduced Gemma 2B and Gemma 7B, smaller open-source AI models designed for language tasks in English. These models, developed by Google, are tailored to cater to simpler tasks such as basic chatbots or summarization needs, offering advantages in terms of speed and cost-efficiency compared to their larger counterparts.
Authored by Emilia David, a journalist specializing in AI coverage and technology, finance, and the economy, the unveiling of Gemma signifies a departure from Google’s previous approach with the Gemini model. While Gemini is a robust closed AI model competing with OpenAI’s ChatGPT, the Gemma models provide developers with more flexibility to leverage Google’s AI technology.
Google emphasizes that Gemma models outperform larger counterparts on crucial benchmarks and can seamlessly operate on personal laptops or desktops. These models will be accessible through platforms like Kaggle, Hugging Face, Nvidia’s NeMo, and Google’s Vertex AI, encouraging broader experimentation and usage.
In line with industry standards, a commercial license will accompany both Gemma model variants, allowing usage across organizations while prohibiting specific applications like weapons development programs. Gemma will also come with “responsible AI toolkits” to maintain ethical standards, offering developers guidelines, restricted word lists, and a model debugging tool for monitoring and rectifying operational issues.
While currently optimized for English language tasks, Google aims to collaborate with the community to expand Gemma’s applicability to other linguistic domains. Developers can access Gemma for free on Kaggle, with incentives such as \\(300 in credits for first-time Google Cloud users and potential grants of up to \\\)500,000 in cloud credits for researchers.
The introduction of Gemma reflects a trend in the AI industry towards compact models, with other companies like Meta launching scaled-down versions of their flagship models. Google’s Gemini lineup includes various configurations such as Gemini Nano, Gemini Pro, and Gemini Ultra, with recent additions like Gemini 1.5 tailored for business users and developers.
Fun fact: The name “Gemma” symbolizes a precious stone, embodying the essence of these innovative AI models.
