top of page
Arrow Graphic.png

An ambitious programme geared to create a radically new architecture for the UK’s internet and telecommunications infrastructure

Microphone Heading.png

BT Thought Leadership series...

A series of linked talks in 2020/2021 by principal academics on the NG-CDI project, hosted by the BT Thought Leadership programme...

Next Generation Converged Digital Infrastructure (NG-CDI)

Professor Nicholas Race.

Network Systems, University of Lancaster. Principal Investigator for NG-CDI.

7th October 2020, from 12.00 - 13.00.

Introduced by Stephen Cassidy. System Science, BT Applied Research

The UK’s Digital Infrastructure is critical to the commercial and social activities and success of the country. It is essential that this infrastructure continues to be world leading. To keep ahead we need an infrastructure which responds quickly to changing needs, and at minimum cost. The creativity of the whole ecosystem will give rise to opportunities that we cannot predict.

This means that services need to be configured in software rather than hardware to reduce the barriers to experimentation and scaling. Distributed autonomic technologies offer the opportunity to manage the expanding scale and complexity, and support faster ways to assess opportunities and risks, make decisions, and simplify service delivery. The operation of such an infrastructure will require new skills, cultures and practices. Nick will introduce and describe the research underway to deliver these aims. He will describe the approaches being taken and how the different aspect of the architecture fit together. 

NR2 (400 x 400).png

Intent Based Networking

Prof. Ning Wang.

Networks, University of Surrey, 5G Centre.

Dr Charalampos Rotsos.

Computer Networks, University of Lancaster.

3rd November 2020, from 13.00 - 14.00.

Introduced by Peter Willis, Software-based Networks, BT Applied Research

Increasing the rate of delivery and value of new services will depend on smarter ways to capture customer needs and translate these into service definition and delivery. The research is investigating the capture of customer intents in machine-readable ways. The research covers not only service creation and DevOps, but also methods to maintain or re-negotiate service levels in real-time in the face of changing network dynamics, using autonomous distributed agent architectures.

HR (400 x 400).png
NW (400 x 400).png

Intelligent Asset Management for Service Assurance & Infrastructure Management

Dr Ajith Parlikad.

Asset Management, University of Cambridge Institute for Manufacturing.

8th December 2020, from 13.00 - 14.00.

Introduced by Arjun Parekh. Self-learning Networks, BT Applied Research

Adding intelligence to network assets offers the possibility that the infrastructure can trigger appropriate maintenance processes. Prognostic maintenance scheduling concentrates engineering effort on reducing the risk to customer service and costs.

Assets can learn from their own experiences, or from swarms of similar assets to anticipate their remaining useful life and co-operatively decide the best means to maintain service, reconfiguring themselves or calling for human help.

Risk models embrace the likely propagation of problems across the regions of the network and between other networks such as the power network. The prospect is offered of determining the best action at the time, based on the dynamics of the existing traffic pattern.

AP (400 x 400).png

Network Assurance through Massive on-line Anomaly Detection

Prof. Idris Eckley.

Mathematics and Statistics, University of Lancaster.

13th January 2021, from 13.00 - 14.00

Introduced by Trevor Burbridge, Diagnostic and Network Assurance, BT Applied Research

New statistical techniques have been developed which are able to monitor very large data streams and identify anomalies in near real time.

Distinguishing between normal randomness and specific types of anomalies is essential for determining when a pattern of events means an intervention is needed. Multivariate techniques extend this to distinguish real system changes from spurious changes in the data.

Such techniques greatly increase the power and discrimination of Test and Diagnostic processes which deliver Network Assurance.

IE (400 x 400).png

Technology, Risk & Organisations

Dr Philip Stiles.

Corporate Governance and Co-Director of the Centre for International Human Resource management, Cambridge Judge Business School.

10th February 2021, from 13.00 - 14.00

Introduced by Stephen Cassidy. System Science, BT Applied Research

Successful operation of an increasingly autonomous infrastructure means that it must connect with the principal business functions of the company. It needs to be trusted at all levels to operate safely. It needs to be driven by well-defined business decisions and goals. It needs to warn us of problems. And it needs to be a vehicle for stimulating innovation, providing information and what-if capabilities to explore new possibilities. These will mean that we need to build increasing trust between human and machine. We need increasing willingness to experiment and comfort with balancing benefits and risks. New systems of governance and support tools which support trust and enlightened risk-taking. Learnings from other industry sectors which are treading these paths, such as Financial algorithmic trading, or Retail logistics, suggests useful approaches.

PS (400 x 400).png

World Models and Digital Networks

In this lecture I will review recent achievements in machine learning underpinned by learning representations in an un(self)supervised paradigm. Such techniques are at the heart of the latest and best performing language models (BERT, GPT-3), contrastive learning computer vision or protein folding predictors (AlphaFold2). The common feature of such techniques is an attempt to build a fundamental understanding of the model i.e. its “world model”, before a subsequent attempt is made to solve the given task. This is in stark contrast to the state-of-the-art techniques (including ML/AI) currently used in digital networks, where the algorithms are specifically crafted to solve the given task(s) and trained from the outset to achieve this. Can digital networks perform better by initially learning their own digital world models?  I will present the case in favour of this view, and not shy from listing arguments against it. 

Prof. Robert Piechocki.

Wireless Systems, University of Bristol, and Turing Fellow.

9th March 2021, from 13.00 - 14.00

Introduced by Arjun Parekh. Self-learning Networks, BT Applied Research

RP (400 x 400).png
Bristol (web).png
Cambridge (web).png
Lancaster (web).png
Surrey (web).png
UKRI (web).png
BT 2 (web).png
bottom of page