Science: The Hardware Infrastructure for AI

The Story:

Nvidia Corp last week announced the new graphic chips (GPUs) that it said will increase the computing speed of the infrastructure for artificial intelligence. This is part of an intensifying competition among chip makers to be the ones creating the hardware for AI, chiefly responding to the lure of data center businesses.

New Chips:

Specifically, Nvidia has unveiled the new H100 chip and a new processor chip that it calls the Grace CPU Superchip, based on technology from Britain’s Arm Ltd. Coincidentally, an effort by Nvidia to purchase Arm fell apart last month.

Nvidia also says that its own new supercomputer, “Eos,” named for the goddess of the dawn, will begin operation later this year, and will be the world’s fastest AI system.

Strange New Worlds:

AI consists of models that generally require to be fed huge amounts of data to attain decent levels of accuracy and plausible speeds.

Companies use AI and machine learning (algorithms that improve automatically through their own experience and data collection) on a wide range of things. This is the tech that tells television viewers what shows they might like to watch, and it is also the tech that helps speed-up the development of pharmaceuticals.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.