Hewlett Packard Enterprise (HPE) is extending the capabilities of its Aruba Networking Central network management platform with new generative artificial intelligence (AI) techniques.

The updated  platform announced today brings in new generative AI (genAI) power to help networking administrators perform tasks like provisioning, monitoring, troubleshooting and more. Aruba Networking Central makes use of Aruba's massive database of historical network data and insights to power the AI.

“The thing is, you're not gonna lose your job to AI, you're gonna lose your job to the person who's able to use AI better than you,” David Hughes, SVP and chief product officer, HPE Aruba Networking said during a press briefing. “So what we want to make sure is that the Aruba admins are getting full advantage of the latest and greatest AI and doing it in a safe way.”

Aruba Networking Central already has some AI capabilities though they predate the modern large language model (LLM) era of generative AI. One of those capabilities is a feature called AI Search.

Alan Ni, senior director, Edge Marketing, HPE Aruba Networking explained that AI Search has already been in Aruba Central for over two years. To date, it has been based on an advanced natural language processing (NLP) technology to help users find answers to networking questions.

“What we've done is effectively take AI search and have done an engine swap and we're taking out some of the NLP engines and we're actually replacing them with large language model transformers,” Ni said.

The net benefit according to Ni is that searching is now faster and more intelligent. He noted that one of the things that LLMs have brought forward is much better success in terms of understanding what a query is all about in terms of intent.

So instead of giving users search results which could have many different line items that require further research by the user, the new AI Search  can bring the users to a specific screen or bring  specific data around that specific question so the user can take effective action.

Ni also emphasized that the new HPE service isn't using ChatGPT or another generic LLM. Rather he said that HPE is using multiple LLMs that have been deliberately built for networking and are not based on just generic internet information. The models have been very specifically trained on a corpus of networking information and questions that HPE has curated over the period of several years.

Network data can have potentially sensitive information and as such, it might be considered a security risk to send information to a third party generative AI service to get insight.

Data privacy and accuracy are key areas of focus for HPE. To help limit any risk, HPE is actually running its own fine-tuned models that are all self-contained within the HPE GreenLake cloud platform. HPE is also implementing a guardrails approach to help limit the risk of hallucination and ensure accuracy.

“This is production grade stuff built for networking,” Ni said.