Google releases AI models that are open-source: Gemma

The new initiative, which comes after similar actions by Meta Platforms and other tech giants, enables both individuals and corporations to design their own AI software using "open models" at no cost.

Additionally, Google is making available important technical data, such as model weights.

With Gemma, the Alphabet-owned firm hopes to increase engagement with its now-profitable cloud segment and attract software developers to Google's platform. Additionally, these models have been optimised for Google Cloud, where $300 in credits are offered as a welcome bonus to new customers.

The computer giant has chosen to keep some control over the conditions and usage of the Gemma models, though, and has chosen not to fully open-source them. This choice is made in the backdrop of continuing discussions over open-source AI's potential for abuse despite its advantages in democratising access to technology.

Google did not expose its flagship Gemini models, which are bigger than Gemma. The business emphasised that the two billion or seven billion parameters used in the development of the Gemma models show the variety of variables an algorithm might utilise to generate outcomes.

In an associated move, Nvidia, a designer of AI chips, announced that it had partnered with Google to guarantee that Gemma models run well on its hardware.

Furthermore, the business announced that it is now working on chatbot software that would work with Gemma and be used to run AI models on Windows PCs.

(Sources: investing.com, reuters.com)