Google continues to offer some artificial intelligence models as open source to developers to expand its work in AI technologies and to provide more collaboration and innovation in this field. Launched in May, the Gemma program hosts some AI models that Google does not use internally. These models are offered openly to developers, thus aiming to make AI technologies accessible and developable by a broader user base.
Google’s New Artificial Intelligence Models
Part of the Gemma program, the Gemma 2 2B model is used for simple text generation and analysis and works across a wide range from laptops to Edge devices. This model is licensed for specific scientific content and commercial applications and can be downloaded through Google’s Vertex AI library.
Another significant model, ShieldGamma, consists of a collection of security classifiers aimed at detecting and blocking hate speech, sexism, and obscene content. Built on the Gemma 2 model, ShieldGamma can effectively filter both commands and responses.
Lastly, the Gemma Scope model, which includes specialized neural networks that allow for more in-depth analysis on the Gemma 2 model and make the inner workings of the model more understandable. This model provides valuable insights to researchers to better understand the information processing and prediction capabilities of Gemma 2. Google’s steps aim to make AI technologies more transparent and accessible, contributing to advancements in this field.