SUNNYVALE, Calif., November 30, 2022 — Jasper, the class-leading AI content platform, and Cerebras Systems, the pioneer in artificial intelligence (AI) computational acceleration, announced a partnership to accelerate the adoption and improve the accuracy of generative AI in enterprise and consumer applications. Using Cerebras’ recently announced Andromeda AI supercomputer, Jasper can train its compute-intensive models in a fraction of the time and extend the reach of generative AI models to the masses.
The rise of generative AI is one of the most significant technological waves in recent history, making it possible to write documents, create images and code software from ordinary text inputs. Based on sophisticated generative AI models, Jasper’s products are used by nearly 100,000 customers worldwide to write copy for marketing, advertisements, books, and more. Using the power of the Cerebras Andromeda supercomputer, Jasper expects to significantly advance AI work, including training GPT networks to scale AI outputs to all levels of user complexity and granularity. final. This improves the contextual accuracy of generative models and will allow Jasper to quickly and easily customize content for multiple customer categories.
“At Jasper, we eliminate the tyranny of the blank page. Our platform provides an AI co-driver allowing creators and businesses to focus on the key elements of their story, not the mundane. The most important thing to us is the quality of the results our users receive. We are extremely focused on continuously adapting our AI models to meet the needs of our customers,” said Dave Rogenmoser, CEO of Jasper. “Our partnership with Cerebras allows us to invent the future of generative AI, doing things that are impractical or simply impossible with traditional infrastructure. Our collaboration with Cerebras is accelerating the potential of generative AI, bringing its benefits to our rapidly growing customer base around the world.
“Our mission is to build the fastest AI hardware for the most complex AI problems. Our recent introduction of Andromeda, the industry’s largest supercomputer, has advanced this vision by enabling our partners to train large language models faster, with greater accuracy, with better results in less time and with less energy,” said Andrew Feldman, Founder and CEO of Cerebras. “Jasper is a company that defines the categories and exact type of customer we’re building for – customers who are trying to transform economies with AI, who aren’t afraid to venture beyond the status quo, and who are interested in doing a job pioneering in inventing new approaches to deliver extraordinary results.
Announced at SC22, the 13.5 million AI-core Andromeda supercomputer delivers cutting-edge performance and near-perfect linear scaling without traditional distributed computing or parallel programming issues, enabling Jasper to design and effectively optimize its next set of models. In initial work, on small workloads Andromeda was faster than 800 GPUs and on large complex workloads it did work that thousands of GPUs were unable to do. In a recent publication on Gordon Bell’s award-winning work, the authors wrote: “We note that for the larger model sizes (2.5B and 25B), training on the 10,240 sequence length data was impossible. on GPU clusters due to -memory errors when calculating attention…. To enable training of the larger models over the full length of the sequence (10,240 tokens), we leveraged AI hardware accelerators such as Cerebras CS-2, both in standalone mode and as an interconnected cluster. It was these GPU-impossible workloads that Jasper began to explore, looking for more accurate and relevant models, trained in less time with less energy.
Andromeda is now available to commercial customers, as well as academics and graduate students. Andromeda is deployed in Santa Clara, CA in 16 racks at Colovore, a leading high-performance data center. The 16 CS-2 systems, with a total of 13.5 million AI-optimized cores, are powered by 284 64-core AMD Epyc Gen 3 x86 processors. The SwarmX fabric, which connects the MemoryX parameter storage solution to the 16 CS-2s, provides over 96.8 terabits of bandwidth. Thanks to gradient accumulation, Andromeda can support all batch sizes, a characteristic that is profoundly different from GPU clusters.
About Cerebras Systems
Cerebras Systems is a team of pioneering computer architects, computer scientists, deep learning researchers, and engineers of all types who have come together to create a new class of computer systems. This system is designed with the sole purpose of accelerating AI and forever changing the future of AI work, enabling customers to accelerate their deep learning work by orders of magnitude.
Founded in 2021, Jasper is an AI content platform that empowers individuals and teams to leverage AI to scale their content strategies. With Jasper, anyone from freelancers to business builders can break writer’s block, repackage what they’ve written, create original imagery, and adapt their content to different formats and languages. Jasper, has been recognized as “one of America’s fastest-growing private companies” by Inc. 5000.
Source: Cerebras Systems
#Cerebras #Systems #Jasper #partner #pioneering #work #generative