Finding what differentiates Cerebras’ AI supercomputer

With state-of-the-art AI computers, Cerebras Systems aims to give scientific research centers and other organizations the opportunity to experiment and see how technology can apply to their goals and business model.

The AI ​​vendor, known for building computer systems for deep learning – on November 14th, introduced Andromeda, a 13.5 million core AI computer. This high-end computer is built with a cluster of 16 Cerebras CS-2 systems and uses vendor MemoryX and SwarmX systems.

Also Read :  Virtual reality experiences aid substance use disorder: Stories: Artificial Intelligence: Key Areas: Research Impact: Indiana University

MemoryX is a memory extension technology that Cerebras will introduce in 2021. It allows the CS-2 to support models with parameters up to 120 trillion. SwarmX, also introduced in 2021, is a contact fabric that extends Cerebras Swarm on-chip fabric to off-chip fabric.

Also Read :  NYC Issues Proposed Regulations Regarding Automated Employment Decision Tools Law // Cooley // Global Law Firm

Andromeda scaled linearly over the load of large language samples using parallel data, distributing data in a parallel computer environment.

Public access

Karl Freund, founder and analyst at Cambrian AI Research, says that as a start-up launched in 2015, Cerebras’s unique feature is that its computers are accessible to the public.

“If you look at all the other competitors who are trying to get their way into this meadow, they are not really preparing anything this big for public access,” Freund said. “.

If you look at all the other competitors trying to make their way into this meadow, they are not really preparing anything big for public access.

Karl FreundFounder and Analyst of Cambrian AI Research

Cerebras competes with other AI hardware vendors, including Nvidia Greater and Graphcore, a startup partner.

And while organizations like Argonne National Laboratory have purchased architecture for supercomputers, they have not done so on a scale.

What Andrew FeldmanThe CEO of Cerebras’ “made here is aware that he needs to invest in building modern computers, giving people access to them, allowing them to find out what they can do with it that they can not do with GPUs and allow The chip goes down. ”Freund continues.

Expenses and partners

Freund said the biggest challenge for Cerebras’ AI computer will be whether it is as easy to use as the vendors claim.

Another challenge is cost.

“We know it is very expensive because this is a very exotic technology,” he added. “But you get a lot for your money.”

Andromeda ‘s first users include Argonne, who is using technology to work on genetic converters, and Jasper, an AI marketing copywriter who is using Andromeda to design its next big AI language model.

Freund said the benefit for vendors like Jasper of using modern AI computers is that they can experiment with large language models without having to spend the money needed to do it themselves. “AI at this scale is still very much in the research phase,” he said.

Daniel Newman, an analyst at Futurum Research, said Cerebras’ AI AI computers also represent the growing trend in the AI ​​hardware market of increasing accelerators.

“They are designed for challenges like what Jasper AI is trying to achieve,” Newman said. “The production of feature-rich chips will continue to be a key component in the next generation of AI development.”



Source

Leave a Reply

Your email address will not be published.