let’s talk about something mind-blowing – ScienceGPT, a trillion-parameter AI powerhouse training at the Aurora supercomputer in Illinois. Get this, it’s fueled by Intel’s Ponte Vecchio GPUs, and yes, the US government is backing it up. Talk about some serious tech cred!
Now, this training gig is no sprint; it’s a marathon. They’re starting with 256 nodes out of 10,000 on the Aurora supercomputer – taking it slow and steady. ScienceGPT is shaping up to be the heavyweight champ of Language Models, leaving even Google’s Bard in the dust. Ogi Brkic, the VP and GM for data center and HPC solutions, spilled the beans, sharing that it’s a powerhouse mixing text, codes, and scientific results to supercharge research.
Picture this: ScienceGPT operating like ChatGPT, but here’s the twist – will it rock different media types? That’s the cliffhanger we’re all waiting for!
Now, let’s talk Aurora – the second exascale supercomputer in US history. It’s making waves on the Top500 list with 60,000 Intel GPUs and 10,000 computing nodes.
