In a significant move to foster innovation in artificial intelligence (AI), Google announced on Wednesday the release of new AI models that outside developers can potentially utilize to create their own applications. This announcement follows a similar initiative by Meta Platforms and others in the tech industry.
The subsidiary of Alphabet, Google’s parent company, revealed that individuals and businesses can now leverage its new family of “open models” dubbed Gemma for free. The company is providing crucial technical data, including model weights, to the public, facilitating the development of AI software.
This strategic move by Google aims to attract software engineers to build on its technology while also promoting the use of its increasingly profitable cloud division. Notably, the models are optimized for Google Cloud, offering first-time cloud customers using them $300 in credits, according to the company’s statement.
However, it’s important to note that while Google is making significant strides towards openness, Gemma is not entirely open-source. This means that the company retains some control over terms of use and ownership. The debate over the advantages and potential abuses of open-source AI continues among experts, with proponents arguing for its ability to democratize AI development and utilization.
Unlike its larger premier models known as Gemini, which remain undisclosed, Gemma models come in sizes of two billion or seven billion parameters. Parameters represent the values considered by an algorithm to generate output. For comparison, Meta’s Llama 2 models range from seven to 70 billion parameters, while Google has not disclosed the size of its largest Gemini models. OpenAI’s GPT-3 model, announced in 2020, boasted 175 billion parameters.
In collaboration with chipmaker Nvidia, Google ensures the smooth operation of Gemma models on its chips. Nvidia also announced plans to integrate its chatbot software, currently being developed to run AI models on Windows PCs, with Gemma in the near future.
