The new Google AI, Gemma 2 is the best of the previous models, like Gemma 1.5, in 27B and 9B parameter sizes. The recent models, Gemma 2 2B, ShieldGemma, and Gemma Scope, are for different use cases but share a safety inclination, unlike Gemma 1.5.
The Gemma series of models is different from Google's Gemini models. Google didn't make the Gemini code usable, but it utilizes Gemini in its products and makes it available for developers. Gemma is a goodwill within the developer community, as Meta wants to do with Llama.
Google declares Gemma 2 2B as a notable expansion of the Gemma 2 family. The Gemma 2 model is thin and used for producing analysis text that can run on different hardware, involving laptops and edge devices. You can download the new member of the Gemma 2 family from sources like the Google Vertex AI model library, Kaggle, or the Google AI Studio toolkit.
Also, Google said their generative AI model transcends all others, like GPT-3.5 models, demonstrating exceptional AI abilities. About ShieldGemma, this is a group of 'safety classifiers' that attempt to discover toxicity like hate speech, harassment, and sexually explicit content. Built on top of Gemma 2, ShieldGemma can filter urges to a generative AI model and its output.
The new Google AI model, Gemma Scope, enables developers to ‘zoom in’ on unique points within a Gemma 2 model. The recent member of the Gemma family is described by Google in a blog post: ‘[Gemma Scope is made up of] specialized neural networks that help us unpack the dense, complex information processed by Gemma 2, expanding it into a form that's easier to analyze and understand. By studying these expanded views, researchers can gain valuable insights into how Gemma 2 identifies patterns, processes information, and ultimately makes predictions.`
The new Google AI models arrived after the United States Commerce Department approved generative AI models in a preliminary report. Like the models from Gemma, Gemma 1.5, and the Gemma 2 family, open models expand generative AI access to smaller companies, researchers, and individual developers. They also highlight the necessity for aptitudes to track these models for problems.
.webp)
