Google has released three new “open” generative AI models, claiming that they are more secure. UNN writes with reference to Tech Crunch.
Details
Google has announced the release of an AI update, which means new “open” models of artificial intelligence that focus on security.
- Additions to the family of generative models Gemma 2 by Google.
The new models, Gemma 2 2B, ShieldGemma, and Gemma Scope, are designed for different applications and use cases, but they share a common focus on security. Google does not provide the Gemini source code, which is used in Google's own products and is also available to developers. Instead, Gemma is Google's attempt to strengthen its “friendly attitude towards the developer community,” the publication writes.
- Gemma 2 2B is a lightweight model for generating text analysis;
It can be downloaded from such sources as Google's Vertex AI model library, Kaggle platform, and Google AI Studio toolkit.
- ShieldGemma is a set of “safety classifiers” that attempt to detect toxicity, such as hate speech, harassment, and explicit sexual content.
Built on top of Gemma 2, ShieldGemma can be used to filter hints to the generative model as well as the content that the model produces.
AMD услід за Nvidia стає компанією ШІ-чіпів - ЗМІ31.07.24, 15:04
- Gemma Scope allows developers to “zoom in” on certain points in the Gemma 2 model and make its inner workings more understandable.
[Gemma Scope consists of] specialized neural networks that help us unpack the dense, complex information processed by Gemma 2, expanding it into a form that is easier to analyze and understand.
By studying these advanced insights, one can understand how Gemma 2 detects patterns, processes information, and ultimately makes predictions, the company informs.
AddendumAddendum
Open models expand the availability of generative artificial intelligence for small companies, researchers, non-profit organizations, and individual developers, according to a report by Google.
Конкурент Google: в OpenAI тестують пошуковику SearchGPT на основі штучного інтелекту26.07.24, 15:40