A recent panel of experts concurred that AI agents are a key element in employing the technology accross various network operations, but emphasised concerns over trust and governance must be addressed before they can be used in closed-loop scenarios.
Reza Rahnama, MD of mobile networks with BT Group, said it views its networks in their entirety and is using the approach to guide AI deployments.
He explained this addresses complexities associated with “creating multiple tools” which do not necessarily possess all of the information required to run its network.
Rahnama is clear AI agents are going to be a powerful tool with the potential to deliver self-optimising networks, but believes it is too early to go fully closed-loop because of the risk of network faults.
Swisscom lead architect for network and infrastructure IT Rudolf Strijkers concurred, explaining he would expect an AI agent to act like a child by “being a little naughty” if allowed the run of the network today. The risks mean the operator currently uses agents “in very restricted cases”.
The executive also sees the longer-term potential for agents, he just believes several elements must be in place first to make the most of them.
Netcracker AVP of strategy Susan White argued consumer LLMs are perhaps skewing the picture of what AI can currently achieve because they “are just probabilistic, they simulate reasoning”.
White believes operators cannot currently “let AI agents on mission-critical aspects” of their networks because they do not understand all of the elements at play or how to fix them.
Even successful attempts to fix one problem could “create ten more faults somewhere else”.
On the up
There was a message of optimism from Liberty Global MD for network and capital allocation Luk Bruynseels, who noted that while limited, “the AI we use today is probably the worst AI that we will ever use”.
He emphasised the rapid pace of change for the technology as a cause for optimism and highlighted AI could ultimately prove to be the solution to many of the challenges it causes.
Bruynseels said the telecoms industry must not be greedy, urging it to “be smart” and share how it puts AI to work in tackling security or data challenges without hampering the evolution of the technology.
Source: Mobile World Live
Image Credit: Stock Image
