brytfmonline

Complete News World

Google launches new open source artificial intelligence tools

Google launches new open source artificial intelligence tools

The Google Coud Next event generated a lot of buzz in the media due to the announcement of AI tools to read emails in Gmail, create and translate captions in Meet meetings in real time, as well as make a summary at the end, and what is most attention-grabbing is Google Vids, Able to help create and edit videos via artificial intelligence.

All these new features, although very interesting, remain in the testing phase and remain unavailable to the general public. But there is another important thing that has not been commented on much, and that is the new open source AI tools that can now be used by anyone with the necessary knowledge and equipment.

These are new AI tools, but they are not new AI tools

There are currently many AI models available, many of which are already open source. Creating new models without significantly different from other models is no longer really relevant. Google's new open source AI tools seem to follow this logic.

Let's start with Max Defusion, it refers to a set of diffusion imaging models written in Python/Jax, such as Stable Diffusion 2, 2.1, and XL. The idea of ​​the project is to help developers during training, modifications and availability of artificial intelligence

Similar is MaxTextwhich combines textual AI models Llama2, Mistral and Gemma, seeking to make it easier to scale implementation on common computers up to large servers, reducing the technical team's work.

Jetstream also collects textual AI models, but in this case, it strives for maximum efficiency when generating results. According to the company Jet Generates 3x more inference per dollar spent compared to other engines like Cloud TPU.

See also  Google uses heavy artillery to counter the ChatGPT threat

Finally, the Optimized-TPU Created in partnership with Hugging Face as an enhanced package for PyTorch, helping you train AI models with ease and greater performance using TPUs.

Speaking of TPU, short for Tensor processing units, it is an application-specific integrated circuit technology developed by Google to accelerate machine learning workloads. The company announced that Google Cloud now has a new version called Cloud TPU v5p, with two to four times better performance than the previous version.

Stay up to date with the latest weekly news about Linux and technology, subscribe to our newsletter!