Daily Newsletter

01 January 1970

Daily Newsletter

01 January 1970

Google gets closer to open-source AI with release of Gemma AI models

Google has taken a tentative step towards making its AI technology open source in contrast with competitors including Meta and IBM which have chosen a wholly collaborative approach.

Kurt Robson February 21 2024

Google has released new AI models that move a step closer to making its proprietary technology open-source.

The move is a change in strategy for the search engine giant, which previously kept its proprietary AI technology private.

However, the company has not been as vocal as Meta and IBM, both of which expressed the need for an “open-science” approach to AI development. 

The search engine giant announced on Wednesday (21 February) that its new family of AI ‘open models’ will be available to experiment with at no cost. The Alphabet-owned company has made the models’ key technical data publicly available. 

Google’s Gemma AI model comes in two sizes: a smaller version for app building and one for customers who wish to create AI software with high-end data centres and chips.

Gemma has been built from the same research and technology used to create Google’s larger Gemini models, the company said in a blog post.

Unlike Meta’s Llama models, which are entirely open-source for research and development, Google has only partially opened up its technology.

The search engine giant remains in charge of choosing ownership and setting terms of use for its Gemma models. 

Google has also refrained from opening up its larger AI models to developers; its main Gemini model, formerly known as Bard, remains private.

Google’s Gemma models are reportedly sized between two and seven billion parameters. This is considerably smaller than Meta’s Llama 2 models, which range between seven and 70 billion parameters in size. 

Parameter is the name given to the variable the AI model uses to make predictions and decisions.

OpenAI’s GPT-3 model is by far the largest of the AI models, reportedly sized at around 175 billion parameters. 

Google has not confirmed the size of its Gemini model. 

Meta and IBM announced a group called the AI Alliance in 2023, which aims to advocate for an ‘open-science’ approach to AI development.

The announcement pits Meta and IBM against other heavyweights in the industry which are creating closed AI systems such as OpenAI and Microsoft.

Uncover your next opportunity with expert reports

Steer your business strategy with key data and insights from our latest market research reports and company profiles. Not ready to buy? Start small by downloading a sample report first.

Newsletters by sectors

close

Sign up to the newsletter: In Brief

Visit our Privacy Policy for more information about our services, how we may use, process and share your personal data, including information of your rights in respect of your personal data and how you can unsubscribe from future marketing communications. Our services are intended for corporate subscribers and you warrant that the email address submitted is your corporate email address.

Thank you for subscribing

View all newsletters from across the GlobalData Media network.

close