Google launched Gemma on Wednesday, new artificial intelligence (AI) models that outside developers can possibly fashion as their own, following a similar move by Meta Platforms and other companies.
Open Models
The Alphabet subsidiary stated that people and organisations can build AI software for free using its new family of “open models” known as Gemma. The corporation stated that it is making essential technical data, such as model weights, publicly available.
Google’s Technology
The move may entice software professionals to build on Google’s technology and promote use of its increasingly profitable cloud segment.
The models are “optimised” for Google Cloud, and first-time cloud customers who use them will receive $300 in credits, the company claimed.
Google did not fully “open source” Gemma, which means the business may still have a say in determining its terms of usage and ownership.
Some experts have warned that open-source AI was ripe for exploitation, while others have praised the strategy for broadening the pool of people who may contribute to and profit from the technology.
In contrast to Gemma, Google did not open up its larger, premium models known as Gemini with the announcement.
It stated that the Gemma models had two billion or seven billion parameters, which refers to the number of alternative values that an algorithm considers when generating output.
Meta’s Llama 2 models have seven to seventy billion parameters. Google hasn’t revealed the size of their largest Gemini models. For comparison, OpenAI’s GPT-3 model, revealed in 2020, contained 175 billion parameters.
Nvidia announced on Wednesday that it has working with Google to guarantee Gemma models run properly on its hardware.
Nvidia also stated that it will soon release chatbot software to run AI models on Windows PCs, in collaboration with Gemma.
To read our blog on “Meta will start labelling AI-generated images,” click here
