Marrakech – Just a week following the debut of its cutting-edge Gemini models, Google isn’t delaying any time in pushing the boundaries of innovation yet again.
Gemini is known as a competitor of OpenAI’s ChatGPT, the notable difference being that Gemini updates its information on a more regular basis according to the search engine.
Designed by Google DeepMind, Gemini is a comprehensive suite of large language models, also known as LLMs, capable of handling text, images, code, and audio seamlessly through a unified user interface.
This time, the tech giant introduces Gemma, a new addition to its lineup.
Inspired by the acclaimed Gemini models, Google’s most recent products, the Gemma 2B and Gemma 7B, stand out as the newest gems in its arsenal.
These models, which are made to serve both business and research needs, hold the potential to open up a world of opportunities.
While Google described its latest models, Gemma 2B and Gemma 7B, as “state-of-the-art,” the absence of a detailed paper on their performance leaves questions lingering.
Google emphasized that Gemma 2B and Gemma 7B follow a decoder-only design, which is similar to the architecture used in its popular Gemini models.
While Google emphasizes that these are open models, it’s crucial to clarify that these models are not open source.
The Google team interprets this flexibility as meaning that developers can use the models for inferencing and fine-tune them as needed.
“Things that previously would have been the remit of extremely large models are now possible with state-of-the-art smaller models,” says Tris Warkentin, Google DeepMind product management director.
Alongside Gemma 2B and Gemma 7B, Google is also launching a revolutionary responsible generative AI toolbox.
Using Gemma, this toolkit seeks to provide developers with the necessary instructions and resources to guarantee the development of safer AI applications.
