Following Meta, Google Releases Open Models of AI

Following Meta Platforms and others, Google on Wednesday released new artificial intelligence (AI) models that third-party developers may be able to customize and brand as their own.
According to the Alphabet subsidiary, anyone can create AI software for free using its new Gemma family of "open models" for individuals and companies. The business said that it is providing important technical information to the public, including what are known as model weights.The change might entice programmers to enhance Google's technology and promote the company's recently successful cloud division. According to the company, the models are "optimized" for Google Cloud, where new cloud users who use them will receive $300 in credits.Although Google did not completely make Gemma "open source," it is possible that the company will continue to have some control over ownership and use restrictions. While some experts have argued that open-source AI is vulnerable to misuse, others have supported the strategy as a means of expanding the pool of potential users and contributors.

Unlike Gemma, Google did not make its larger, premium models, known as Gemini, available with the announcement. According to the article, the Gemma models have two billion or seven billion parameters, or the total number of possible values that an algorithm considers before producing an output.
The size of the Llama 2 models from Meta ranges from seven to 70 billion parameters. Google has not disclosed the size of its largest Gemini models. For comparison, OpenAI's GPT-3 model announced in 2020 had 175 billion parameters.Chipmaker Nvidia said on Wednesday that it has collaborated with Google to guarantee that Gemma models function properly on its hardware. Additionally, Nvidia announced that Gemma will soon be compatible with chatbot software, which it is currently developing to run AI models on Windows PCs.