Google unveiled a new open-source artificial intelligence model called Gamma yesterday, Wednesday, allowing external developers to design their own models. This move follows a similar step by Meta and others, according to Reuters.
Alphabet, Google’s parent company, stated, “Individuals and companies can create AI programs based on the new Gamma model from the family of open models.” The company clarified that core technical data, such as model weights, is available to the public.
This initiative is expected to attract software engineers to leverage Google’s technology and encourage the use of its cloud division. The company mentioned that Gamma models are “optimized” for Google Cloud, where first-time service users receive a $300 credit.
Google has not made Gamma completely open-source, meaning the company may still have a role in defining usage terms and ownership. Gamma will allow developers to create their own guidelines or list of prohibited words when integrating them into their projects. It also includes a tool for error correction, enabling users to investigate Gamma’s behavior and address issues.
According to The Verge, these models will work best with tasks related to the English language for now. Google hopes to collaborate with developers and the AI community to meet the needs of other language-speaking markets.
Google has not opened its larger and distinguished models known as GPT (Generative Pre-trained Transformer) like Gamma. The size of Gamma models ranges from 2 to 7 billion parameters, which are the different values the algorithm considers to generate outputs.
Meta’s AI models range from 7 to 70 billion parameters in size, and Google has not revealed the size of its larger GPT models, known as Gemini.
NVIDIA announced yesterday that it worked with Google to ensure smooth operation of Gamma models on its chips. NVIDIA also mentioned that it will soon make its chatbot software compatible with Gamma, enabling it to run AI models on Windows-based computers.
Leave a Reply