Google Releases Open AI Models Focused on Security
Google has released three new AI models, which it claims are safer and more transparent.
Google continues its efforts to make AI tools open source. The Gemma program, launched in May, includes models that Google does not use as source code for Gemini, and these models are also available to developers.
Recently, the U.S. Senate indicated that measures would be taken to prevent monopolies in artificial intelligence. Google’s announcement suggests that the company wants to ease its position in the spread of AI research.
Here are Google’s new models.
Gemma 2 2B is a simple text generation and analysis model that can run on nearly any hardware, from laptops to edge devices. This model is licensed for specific scientific content and commercial applications and can be downloaded from sources like Google’s Vertex AI library.
ShieldGamma is described as a collection of “security classifiers.” The purpose of this model is to block hate speech, sexism, and obscene content. Built on Gemma 2, this model can filter both commands and responses.
Finally, Gemma Scope is a model that allows developers to focus on specific points within the Gemma 2 model and make its inner workings more interpretable. According to Google’s blog, this model… “Gemma Scope consists of specialized neural networks that help us unveil the dense, complex information processed by Gemma 2 and expand this information into a form that is easier to analyze and understand. Researchers can examine these expanded views to gain valuable insights into how Gemma 2 identifies patterns, processes information, and ultimately makes predictions,” as explained on Google’s blog.
