Thursday , April 22 2021

University of St. Gallen receives the most powerful computer in the world



St. Gallen is the first Swiss university to receive a supercomputer and is currently considered the world's most powerful artificial intelligence computing device.

Bruno Knellwolf

St. Gallen University has discovered an artificial intelligence research supercomputer. The media interest is great. (Image: Beat Belser)

St. Gallen University has discovered an artificial intelligence research supercomputer. The media interest is great. (Image: Beat Belser)

There is a package from Taiwan in the middle of the room. This is a supercomputer. One that did not exist until yesterday in Switzerland. And if a calculator with the same performance was displayed five years ago, this computer would have filled the entire space.

But the supercomputer named Nvidia DGX 2 is only about a meter long and weighs 196 kilograms. It is believed to be the world's most powerful computing machine for contemporary artificial intelligence (AI) processes.

Thus, Damian Borth, professor of artificial intelligence and machine translation, is happy to spy on a supercomputer. For two months, he has created a new computer science program in St. Petersburg. Gallen University. With Nvidia DGX 2, he has a tool in his hand, unlike any other Swiss university.

Half of the work will be done soon by the machines

The plant believes Borth is a turning point in artificial intelligence research. And it more and more determines our daily lives. By 2025, more than half of the work is expected to be done by machines. AI is becoming an integral part of our pop culture and, in parallel, it becomes increasingly clear that there are too few AI experts and only 22,000 worldwide. Borth wants to help.

Incidentally, with this supercomputer, which has two Petaflops computing power, as Ralph Hintche says about the manufacturer of Nvidia. This is an incredibly large number, so he offers a comparison:

If everyone on earth, seven billion, had 300,000 advanced calculators and all pressed the function in the same second half, it would be a force for two petaflops.

You will need this feat for the artificial intelligence sub-section "Deep learning". It trains the human brain as many models as possible for neural networks. Deep training with artificial nets is an effective continuous machine learning technique. It works by performing a huge amount of statistical analysis. St. Gallen University's two large datasets are handled with a $ 399,000 supercomputer.

AI systems are trained rather than programmed

Damian Borth will work with a great deal of imagery, including from the ESA ESA. For example, an image recognition machine using neural networks causes a self-driving car to determine when a child jumps in front of the car and the vehicle learns to respond properly. Professor Siegfried Gimp will use Nvidia's DGX 2 voice and text analysis. With entries from the social media. The computer will learn to understand the language. "Special Swiss Switzerland is a challenge to artificial intelligence," says the glove.

Applications for AI exist, for example, in robots in patients with dementia. The robot becomes a friend who learns to understand the patient himself. "We do not design AI systems here, we teach it." It should quickly facilitate scientific publications on machine translation and, secondly, facilitate the creation of AIs.


Source link