Google has released the Gemma family of open models, which is a step forward in deepening Google’s commitment to open technology while also maintaining a safe and responsible approach. One of the advantages of open source is that its license grants users full creative autonomy, which ensures technology access for developers and end-users. However, when in the hands of a malicious actor, the lack of restrictions can also pose risks.
The Gemma models are being released as “open models,” with free access to the model weights. Still, the terms of use, redistribution, and variant ownership vary according to each model’s specific terms of use. These models are freely available to individual developers, researchers, and commercial users for access and redistribution.
The approach involved a systematic internal review following the AI Principles, where models were only released when Google determined that the benefits outweighed the risks of misuse. Furthermore, Google is releasing a Responsible Generative AI Toolkit for developers as it releases the Gemma models, providing guidance and tools to help them create safer AI applications.
Source: Google Open Source
Leave a Reply