Google on Wednesday released a new artificial intelligence (AI) model that outside developers can adopt as their own.
Alphabet, Google’s parent company, said that individuals and companies can develop artificial intelligence software based on a new family of open models called Gemma, free of charge. Claiming that it makes important technical data, such as model weights, available to the public.
This initiative could encourage software developers to build on Google’s technology and take advantage of the lucrative new cloud computing sector.
The models are optimized for Google Cloud, and the first customers of this service will receive a $300 credit, the company said.
Some experts believe that open-source AI is ripe for abuse, while others argue that the approach is to widen the circle of people who can contribute to and benefit from the technology.
Comparison with existing models
In this announcement, unlike Gemma, Google is not releasing a larger flagship model called Gemini. It specifies that the parameter size of the Gemma model is between 2 and 7 billion, the number of different values the algorithm takes into account to produce the result.
Metas Llama 2 models range in size from 7 to 70 billion parameters. Google has not revealed the size of its largest Gemini model. By comparison, OpenAI’s GPT-3 model released in 2020 had 175 billion parameters.