SAN FRANCISCO: Google on Wednesday released new artificial intelligence (AI) models that outside developers can potentially create as their own, following a similar move by Meta Platforms ( META.O ), opens a new tab and more.
The Alphabet ( GOOGL.O ) is opening a new branch on the card, which said individuals and businesses can create AI software for free based on its new family of “open models” called Gemma. The company makes key technical data, such as so-called model scales, publicly available.
The move may attract software engineers to build on Google’s technology and support the use of its newly profitable cloud division. The models are “optimized” for Google Cloud, where customers using the cloud will receive $300 in credits for the first time, the company said.
Google has stopped making Gemma fully “open source,” meaning that the company can still have a hand in setting the terms of use and ownership. Some experts said open-source AI was ripe for abuse, while others championed the approach of expanding the pool of people who can contribute to and benefit from the technology.
With the announcement, Google, unlike the Gemma, did not unveil its larger, premier models known as Gemini. He said Gemma’s models are scaled to two billion or seven billion parameters — or the number of different values the algorithm considers when generating the output.
Meta’s Llama 2 models range in size from seven to 70 billion parameters. Google has not disclosed the size of its largest Gemini models. In comparison, the OpenAI GPT-3 model announced in 2020 had 175 billion parameters.
Chipmaker Nvidia ( NVDA.O ) is opening a new tab on Wednesday and said it has worked with Google to ensure Gemma models run smoothly on its chips. Nvidia also said it will soon introduce the chatbot software it develops to run artificial intelligence models on Windows PCs to work with Gemma.
