HPE is introducing two new systems to accelerate and simplify artificial intelligence and a system to do remote machine learning and transfer that data in a privacy-friendly way.
Swarm Learning is a solution for edge and distributed sites. Containers that integrate smoothly with the HPE swarm API with AI models, the company says. The big advantage of machine learning at the edge is that large data volumes do not have to be moved or copied.
It must also be possible to share what is learned, internally or externally, without accessing the source data. Think of analyzing CT or MRI scans to recognize certain things automatically, but without anyone having access to the individual scans afterwards.
Remarkably, HPE says it uses blockchain to onboard members into such projects and must sell that only the data that the machines teach themselves can be shared and not the source data itself. What that has to do with blockchain is not clear to us, but undoubtedly a marketer at HPE thought ‘blockchain’ still sounded hip enough to mention.
In addition, the company also comes with its HPE Machine Learning Development System. Basically a machine based on the HPE Apollo 6500 Gen10 specifically for machine learning. Those machines come standard with at least eight 80GB Nvidia A100 Tensor Core GPUs. The new machine comes further from the acquisition of Determined AI.
HPE promises an end-to-end solution that brings together the machine learning software platform, computing power, accelerators, and networking to train AI models on a large scale accurately. In addition, the company claims to be faster than ‘a cloud provider’ with 32 GPUs, without providing further details.