Google is continuing its efforts in the field of artificial intelligence. The Mountain View firm unveiled Gemini 1.5 with some great new features, but that’s not all. The company is launching open source AI models called Gemma.
Optimized for Nvidia GPUs and Google Cloud
Gemma 2B and Gemma 7B, respectively equipped with two and seven billion parameters, allow individuals and businesses to create AI applications. Behind open source lies a simple objective: to push users to the Google cloud. New Google Cloud sign-ups receive $300 credit for these frameworks that deliver better performance from within the service. A credit of up to $500,000 may be received by researchers (a form has been posted online to register).
AI is also accessible via the Kaggle platform and in Google Colab, which allows you to write and execute Python code from a browser.
Open source, but not completely
Gemma 2B and Gemma 7B are “pre-trained and adapted to instructions†to work much better on PCs equipped with an Nvidia GPU, as well as on Google Cloud TPUs. The Mountain View firm claims that its models are smaller than others, and even outperform “references†in this matter, like Llama 2 by Meta.
Gemma is certainly an open source model, but Google retains its authority over the conditions of use. To avoid deviations, the company imposes precise rules. Because if technology seeks to improve our lives, particularly in the field of health, there are risks constantly discussed by experts and governments.