Aurora genAI is a new generative artificial intelligence model developed by Intel specifically for scientific purposes. It was presented at ISC, the technology fair currently taking place in Hamburg, Germany.

After announcing the supercomputer Aurora, Intel is also introducing a new model for generative artificial intelligence: Aurora genAI. It was specially developed for scientific purposes and has the ambitious goal of containing no less than a trillion parameters. The supercomputer Aurora will be used, among other things, to run Aurora’s AI model.

Scientifically recognized

Aurora genAI’s focus is really purely scientific. The model is trained on both general and scientific text, scientific data, and related code.

Possible applications are:

• systems biology
• cancer research
• climate sciences
• cosmology
• Polymer chemistry and related materials
• General Sciences

The foundations Aurora genAI is built on are Megatron, Nvidia’s machine learning model, and the open-source DeepSpeed ​​optimization library. The ultimate goal of these trillion parameters requires a lot of training. To put that in perspective, the free and public version of the popular ChatGPT targets 175 million parameters…

Meanwhile, in other AI news, Intel confirmed that the Falcon Shores GPUs – intended for large-language models – will be ready in 2025. The plans for the “superchip” Falcon Shores XPU were finally put on hold.