Listen to the audio version of the article
Google doesn’t stop and continues to amaze with its innovations in the AI field. After announcing Gemini Advanced a few days ago, the “Ultra” artificial intelligence model capable of carrying out highly complex tasks, and Gemini 1.5, available for companies, today it makes Gemma open source models accessible to all, which can be used by developers external users to create their own AI models.
This is how Gemma is presented on the official blog: “Developed by Google DeepMind and other Google teams, Gemma is inspired by Gemini and the name reflects the Latin gemma, which means ‘precious stone’. Along with our model weights, we are also releasing tools to support developer innovation, promote collaboration, and guide the responsible use of Gemma models.”
The release is made available in two formats: Gemma 2B and Gemma 7B, with 2 and 7 billion parameters respectively. These are the same models used by the company to create Gemini, with which they share technical and infrastructural components.
What can be done with Gemma
Following a similar strategy of Meta, which had made available some basic codes necessary to develop chatbots, Google’s open source models offer individuals and companies the possibility of developing their own artificial intelligence applications for free and with the advantage of being able to run directly on the developer’s computers and offering performance comparable to or superior to that of larger models, such as Meta’s Llama 70B or Mistral AI’s Mistral. In a market that promises to be very crowded, Google hopes to convince developers to use and test its technologies and to rely on its cloud services, where newcomers are offered $300 in service credits. A strategy that also includes a partnership with the semiconductor manufacturer for artificial intelligence Nvidia, which will guarantee the proper functioning of the Gemma models on its equipment.
Is the future of AI open source?
Many experts believe that the future of AI is open source. According to the 2023 Open Source AI Report published by the Linux Foundation, 41% of companies surveyed prefer open source generative AI technologies over proprietary ones, as they offer benefits such as speed, privacy, capacity and collaboration. The open source AI “war” has just begun and involves all the big technological giants, in addition to Meta and Google there are Alibaba, Amazon and IBM, who have already launched their proprietary AI platforms. Each of them aims to attract developers, offering them high-quality models and encouraging the use of their services and platforms. Google is no exception and with Gemma it demonstrates its desire to be a protagonist in this competitive and innovative scenario.