Google launches new versions of Gemma 2 AI with a focus on security

byTecYLife -
0

gemma 2,gemma 2 2b,gemma 2b,gemma 2 google ai,shieldgemma,gemma scope,google ai,google gemma 2,gemma 2 release,google gemma,gemma 2 features,gemma 2 ai model,how does gemma 2 work,gemma,google machine learning,ai security,google,gemma 2 capabilities,how to use gemma 2,gemma 2 abilities,google's gemma 2 ai

Google announced three new additions to the Gemma 2 family of artificial intelligence models, focusing on security and transparency. The latest models, Gemma 2 2B, ShieldGemma, and Gemma Scope, are designed to offer robust yet affordable solutions for developers and researchers.

"With these additions, researchers and developers can now create safer customer experiences, gain unprecedented insights into our models, and confidently and responsibly deploy powerful AI directly on the device, unlocking new possibilities for innovation," the company highlighted in its announcement. Wednesday (31).

New versions of Gemma 2

The new versions are derived from the Gemma 2 open model, launched in June in two sizes: 27 billion and 9 billion parameters. See details of the latest models:

Gemma 2 2B

Gemma 2 2B is the 2 billion parameter version of the Gemma 2 model that can run on different types of hardware, including laptops and high-end devices. It is available for some commercial and research applications and can be downloaded from sources such as Google's Vertex AI model library, data science platform Kaggle, and Google's AI Studio toolkit.

gemma 2 2b,gemma 2b,gemma 2,gemma 2 google ai,gemma scope,google ai,google gemma 2,gemma 2 release,google gemma,gemma 2 features,gemma 2 capabilities,gemma 2 ai model,how does gemma 2 work,gemma,google machine learning,ai security,gemma 2 abilities,google's gemma 2 ai

"This lightweight model produces outsized results by learning from larger models through a distillation process. Gemma 2 2B outperforms all GPT-3.5 models and demonstrates exceptional conversational AI abilities," highlights Google in the statement.

ShieldGemma

shieldgemma,gemma 2,gemma 2 google ai,gemma scope,google ai,google gemma 2,gemma 2 release,google gemma,gemma 2 features,gemma 2 capabilities,gemma 2 ai model,gemma,google machine learning,ai security,how to use gemma 2,gemma 2 abilities,google's gemma 2 ai

ShieldGemma is a suite of security classifiers designed to filter the inputs and outputs of AI models to ensure a safe user experience. The classifiers can detect and mitigate harmful content such as hate speech, harassment, and sexually explicit content.

Gemma Scope

Gemma Scope is a tool that offers in-depth insight into how Gemma 2 models work. "Acting like a powerful microscope, Gemma Scope uses sparse autoencoders (SAEs) to magnify specific points within the model and make its inner workings more interpretable," explains Big Tech.

These autoencoders translate the information processed by Gemma 2 into a format that is 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," says Google.

According to the company, Gemma Scope can help the AI ​​research community "discover how to build more understandable, accountable, and trustworthy AI systems." The model is available on the Neuronpedia platform.

 

Tags

You may like these posts

Previous Post Next Post

نموذج الاتصال