IBM Corp. is teaming up with the artificial intelligence startup H2O.ai Inc. in an effort to speed up the training of machine learning models by using its latest POWER Systems servers.

The companies today announced a partnership that involves running H2O’s Driverless AI platform atop IBM’s servers, which have been fine-tuned for training and running machine learning models.

H2O designed Driverless AI to automate much of the complex and repetitive work that’s involved in AI development. Driverless AI works by helping users to choose the best machine learning model for their particular task.

It then customizes that model for the specific type of data being processed. So, for example, if a bank is building an AI system for spotting fraudulent transactions, it would prioritize the kinds of variables that indicate this kind of activity with its training.

IBM’s POWER 9 systems, meanwhile, were built specifically for compute-intensive AI workloads, which is why the partnership makes sense. In a blog post, Sumit Gupta, IBM’s vice president of AI and machine learning, said the POWER 9 systems can boost machine learning training times by four times that of regular servers, thanks to a combination of PCI-Express 4.0, NVIDIA NVLink and OpenCAPI technologies that help to accelerate the movement of data.

IBM said Driverless AI on POWER Systems should be of particular interest to companies in heavily regulated industries such as financial services, where organizations need to be able to understand how AI technology reaches its conclusions.

“Just as children often aren’t satisfied with parents telling them they can’t do something ‘because they said so,’ so to are enterprises needing additional justification for changing business processes because an algorithm said so,” IBM officials said.

Driverless AI helps with this by tracking the development of machine learning models and helping organizations to understand why it makes specific decisions or predictions over others. As a result, it should help to open up potential use cases in regulated industries where greater confidence in decision-making is needed, IBM said.

Image: geralt/pixabay