During the last few years, we have experienced a real expansion in artificial intelligence, which has even surpassed the Turing Test. However, America had been working even longer on an invention that could put an end to it. What have we achieved? A group of experts has turned on the “machine of the future”, with the capacity to process a quintillion operations per second, and to revolutionize the whole of humanity in this way.
America has fired up the “machine of the future”: It could be the end of artificial intelligence worldwide
Oak Ridge National Laboratory (ORNL) has yet again set a record with Frontier, the world’s fastest supercomputer. Frontier is going to bring a new era of computing capability to solve complex problems in various domains and open up a host of new areas in the field of AI. Our purpose is to elaborate on why Frontier is a groundbreaking advancement in AI, or rather, the end of artificial intelligence.
This marks the start of the true AI age, wherein Frontier delivers as the world’s first exascale supercomputer, executing more than quintillion calculations every second. The main feature of BigScience is its computational capability, which makes it possible to train large language models that were not attainable by many researchers.
Frontier democratizes powerful AI tools to ignite the possibilities of scientific discovery and innovation for a broad range of users. Frontier is really a masterpiece of engineering; the design of this appliance is significantly different from your run-of-the-mill supercomputers. In essence, Frontier operates with the help of over 9,400 nodes.
Every node is equipped with four AMD MI250X graphics processing units. This GPU-centric design allows Frontier to perform well in both traditional high-performance computing and new-age AI workloads.
6.8 quintillion operations per second: How Frontier works (and why experts cannot believe it)
These are general performance figures that are quite impressive, to say the least, Frontier’s. In mixed-precision computing, which is widely used in AI and machine learning, the performance of Frontier hit 6.88 exaflops, or 6.8 quintillion calculations per second (something that artificial intelligence has never achieved).
This unprecedented speed enables the researchers to solve problems that were believed to be unsolvable before. Likewise, Frontier has outstanding full-precision performance, which is essential for classical HPC simulations. The supercomputer obtained a peak performance of 1.9 exaflops; this result is 10 times more promising than in the ORNL’s previous summit.
This level of performance allows scientists to solve the issues on a national scale, like cancer, climate change, or nuclear fusion. Despite this, the establishment of Frontier came with its own set of difficulties. There were several technical challenges that ORNL’s researchers had to overcome in order to build a machine for exascale computing.
Experts have an even better software: How it could be used in laboratories across America
However, it wasn’t just the physical characteristics of Frontier that received such treatment; even the software was optimized to the best of the Possibilities ASF, along with the developers of the cooling system, thought was given to every possibility of improving Frontier in any way.
Frontier’s imprint on the world of AI has started to make its way into the realm of reality. Scientists from GE Aviation and GE Power are among those who have started using the potential of Frontier to create hydrogen propulsion and hybrid electric systems. This synergy exemplifies how Frontier might spearhead innovation across areas from space to energy.
As you can see, the days of artificial intelligence could be numbered, now that we have managed to overcome them with machines based on computational thinking and supercomputers. This software directly rivals two other very similar ones, both in power and programming, which have been developed by China and Japan separately. However, in America, we hold the key to pushing it to a new limit and shattering all AI forecasts for the next few years.
