HPE and Aleph Alpha: Advancing the AI Revolution

2023 SEASON SALE Networking and Security Showcase In-stock ICT products at exclusive discounts

One hundred thousand years ago, the birth of the first language marked the different paths of humans and apes. This moment reveals the era when human society moved from barbarism to civilization. However, even after 100,000 years, the “evolution” of the language continues. In today’s era of rapid development of artificial intelligence, how to break through the communication barriers between different languages has become an urgent problem for human beings to solve. Aleph Alpha from Germany focuses on the development of natural language processing models and cooperates with HPE to break through language barriers and promote technological progress.
There is no doubt that artificial intelligence will bring about huge social changes. One of the key challenges is linguistic diversity. AI inputs and outputs need to be equally accurate and complete. Given the continent’s linguistic diversity, a better understanding of language and breaking down communication barriers between different languages are prerequisites for the development of artificial intelligence. Aleph Alpha, a start-up company from Germany, is fully aware of this and has participated in the European Artificial Intelligence Beacon Project. With the support of HPE, Aleph Alpha successfully developed the Luminous large-scale language model, which has been able to support five major languages including English, French, German, Italian and Spanish.

Provide artificial intelligence as a service based on HPE technology

To train a new generation of natural language processing models, Aleph Alpha has built one of the most powerful data centers in the world to manage huge high-performance computing workloads. The new data center is not only powerful, but also very smart. Aleph Alpha can isolate different workloads for different projects and users, and it requires far less water and energy to run the facility than a standard data center.
The data center is based on HPE technology, using HPE software, hardware and services, and completed all the work in less than six months. Lutz Kirchner said, “For us and our AI customers, turning an empty room into a data center in just over six months is incredible, and this level of power and production capacity is fantastic. HPE has indeed provided us with a lot of help, and customized a series of equipment for us, which no one except HPE can do.”

HPE’s Industry-Leading Hardware System Acceleration Insights

The data center is equipped with HPE Apollo 6500 Gen10 Plus systems with 80GB NVIDIA® A100 GPUs and HPE ProLiant DL385 servers with AMD EPYC processors, the world’s highest performing x86 server CPUs. Powerful hardware support allows Aleph Alpha to generate 800GB of data across 64 nodes, and supports Aleph Alpha to configure CPU, memory, and storage according to use cases. Additionally, the solution leverages the HPE Machine Learning Development Environment, which is built on an open-source training platform. Aleph Alpha therefore eliminates the need to write infrastructure code, which not only speeds up production, but also makes it easy to set up, manage, secure and share AI computing clusters.
“It’s so fast and flexible, we can automate a variety of different workloads in a matter of minutes, and the AMD processor fits perfectly with the whole system,” Kirchner said. Our request is that there will be no power outages.”
Also featured in the solution is the HPE Adaptive Rack Cooling System (ARCS), which maximizes computing power without increasing the thermal load. Kirchner explained, “I was initially skeptical about HPE ARCS, but the results blew my mind. Each ARCS rack draws 50 kilowatts of power, compared to the normal 15 kilowatts per rack. This means the A data center of 200 square meters has the performance of 2000 square meters. It can be said that the excellent performance brought by HPE and AMD makes us more competitive.”