VIAVI: The decomposable AI landscape

A reality check on AI, ML and GenAI in telecom with Per Kangru and Takai Eddine Kennouche at VIAVI Solutions

AI will change the way we operate telecom networks. But what’s the best way to get there? How should we introduce AI to get the benefits it promises but avoid duplication, conflicts among different AI models, and undesired outcomes? How do we know AI models will deliver what we expect?

Per Kangru and Takai Eddine Kennouche at VIAVI Solutions shared their AIOps practical experience and vision for integrating AI and ML in network operations in an efficient and streamlined way while preserving the reliability and quality of experience that are non-negotiable.

Much of the discussion was about how a decomposable AI approach allows operators to use a common framework where data and models can be used for multiple tasks and thus reduce costs, duplication and energy consumption, thus strengthening the business case for AIOps to deliver a higher operational efficiency.

Conversation timestamps

  • 00:00 Introduction
  • 05:42 Per and Takai introduce themselves
  • 07:23 Hype and reality, expectations
  • 09:44 What operators need
  • 12:11 What it takes to get AI to work, foundations, reliability; talent and technology; scalability
  • 16:46 ROI for AI models and specific use cases, where does the value come from?
  • 19:39 Composable AI, customization, reuse, domain-specific approaches; open-source, interoperability; CI/CD
  • 23:21 Use cases, sustainability and energy efficiency, AI is good at what humans are not good at
  • 29:30 RIC, xApps, rApps
  • 32:39 Feeding the right data to the models, network architecture, RIC and Open RAN, generating data to feed to the models
  • 46:24 Cultural change driven by AI
  • 51:18 Customer feedback; need to talk about the value of AI; more on use cases, more on Open RAN
  • 58:35 Role of hyperscaler, AI lock-in, regulation, GenAI, adversarial learning
  • 1:03:51 How to adopt decomposable AI

Register to choose which sections to watch (or read)

Register to view or read the conversation

* Required

We use cookies to improve your browsing experience. By continuing to use the website you allow Senza Fili to deploy cookies, as detailed in our privacy policy.