Google has unveiled its new family of "open models" called Gemma, allowing outside developers to potentially create their own AI software. This move, similar to initiatives by Meta Platforms, enables individuals and businesses to build AI applications for free, with key technical data, including model weights, made publicly available.
While not fully "open source," this approach aims to attract software engineers and promote the use of Google's cloud division, offering optimized models for Google Cloud with $300 credits for first-time users.
Unlike the premier Gemini models, Gemma models come in sizes of two billion or seven billion parameters, indicating the values considered by the algorithm to generate output.
Google's decision not to make Gemini fully open raises discussions about terms of use and ownership. While some experts express concerns about potential abuse of open-source AI, others see it as a means to broaden contributions and benefits from the technology.
Meta's Llama 2 models range from seven to 70 billion parameters, and Google's largest Gemini models remain undisclosed. In comparison, OpenAI's GPT-3 model from 2020 featured 175 billion parameters.
Chipmaker Nvidia has collaborated with Google to ensure Gemma models run smoothly on its chips, and they plan to make chatbot software compatible with Gemma for AI models on Windows PCs.
0 Comments