| 27 June 2024 | The Computer Laboratory, University of Cambridge | |
| 09:00 - 09:30 |
Registration | Exhibition & Networking |
| 09:30 - 09:40 |
Welcome from Cambridge Wireless Michaela Eschbach, CEO, Cambridge Wireless
Welcome from our host Ian Wassell, Senior Lecturer, Department of Computer Science & Technology, University of Cambridge |
| 09:40 - 10:10 |
Introduction to AI Chair: Michaela Eschbach, CEO, Cambridge Wireless This session outlines AI basics, its evolution, and the current surge in technologies like convolutional neural network (CNN) and Generative AI, setting the stage for understanding AI's role in modern applications.
What is AI? Mary-Ann Claridge, Founder, Mandrel Systems Phil Claridge, Founder, Mandrel Systems
|
| 10:10 - 10:35 |
Keynote: Learning to Communicate, Cooperate, and Coordinate in Multi-Robot Systems How are we to orchestrate large teams of agents? How do we distil global goals into local robot policies? Machine learning has revolutionized the way in which we address these questions by enabling us to automatically synthesize decentralized agent policies from global objectives. In this presentation, I first describe how we leverage data-driven approaches to learn interaction strategies that lead to coordinated and cooperative behaviours. I will introduce our work on Graph Neural Networks, and show how we use such architectures to learn multi-agent policies through differentiable communications channels. I will present some of our results on cooperative perception, coordinated path planning, and close-proximity quadrotor flight. |
| 10:35 - 11:05 |
Break | Exhibition & Networking |
| 11:05 - 12:45 |
Applications of AI in Critical Industries Chair: Paul O'brien, Independent Consultant This session explores AI's impact across telecommunications, cybersecurity, healthcare, and banking, showcasing its capacity to enhance infrastructure resilience, cybersecurity defences, healthcare diagnostics, and banking services. We will explore a selection of use cases where AI has and is being applied to critical industries.
Chris Murphy, Regional CTO, EMEA, Viavi Solutions
Stephen Clemmet, CEO, Silogic Technology The Future Of Wind Turbine Cybersecurity This talk will explore using AI to detect anomalies in SCADA data for national critical infrastructure. The focus will be on machine learning to detect live cyber-threats, and related operational performance on wind turbines. The story to be shared with pictures examines how AI would have detected a two-stage attack, the first stage was a denial-of-service, and the second stage was an on-turbine attack. The asset owner didn't know about the attack or want to pay the ransom demand, so the attack lasted for six months. Whereas, if they had AI, it would have been picked up within a day and AI would have told them what was being attacked during the on-turbine attack.
Dr Camille Terfve, Partner, Mewburn Ellis AI in health, pharma and biotech Life sciences have been data driven for a while now: from omics and in vitro imaging to healthcare records via medical imaging and sensor data, there is an absolute wealth of noisy information to work with at every scale. This means that the opportunities for application of AI are equally abundant. This talk will provide a snapshot of the breadth of application of AI in the health, pharma and biotech sector, in terms of problems tackled (e.g. diagnosis, drug design, patient monitoring, etc.), data used (e.g. imaging, omics, etc.) and types of AI used (e.g. simple classifiers to large language models). We will also briefly touch on the specific challenges of working with AI in these fields.
Cheryl Allebrand, Senior Consultant, Artificial intelligence (AI) and Automation, CGI Eyes on transport: Machine Vision support of critical services on the Edge According to Network Rail, scour is the leading cause of bridge failures in the last 100 years in the UK and 4500 structures are currently at risk of scour, of which 750 are rated as High or Medium/High risk. Despite best efforts, sending divers to monitor erosion doesn't match the speed with which erosion develops, and is ambiguous compared to the level of intel that real-time sensors and imagery can provide. This brief talk will touch on:
|
| 12:45 - 13:45 |
Lunch | Exhibition & Networking |
| 13:45 - 15:15 |
AI Engineering Challenges Chair: Peter Whale, Founder & CEO, Vision Formers This session discusses the practical aspects of building AI solutions, including data management, risk mitigation, interoperability standards, and deployment tools, emphasising the organisational readiness required for AI integration. Focusing on the practice of engineering AI solutions and the challenges for organisations to become AI Ready, including business outcomes and technical results from deployment.
Simon Thompson, Head of AI, ML & Data Science, GFT Financial Ltd. A secure and useful Banking application built with LLM’s? Managing personal finances can be complicated and confusing for many consumers. Retail banks want to be differentiated and target customers with products that are tailored to them but struggle to get their message across. Using an LLM to create an intelligent banking application might allow a greater level of customization and control for customers and create a new channel for retail banks. The challenge is that LLM’s are insecure and unreliable. In this talk I will describe an intelligent banking app that leverages LLM’s and describe the tactics that were used to put it together, as well as the limitations that we encountered.
Objective quality, performance and cost evaluations of AI solutions Adopting cutting-edge AI solutions in critical industries such as banking leads to numerous engineering challenges: from strict regulatory requirements to high development and deployment costs. Amidst constant innovation in hardware, software and "neuralware", it's becoming ever more difficult for practitioners to evaluate and select among many possible options and a myriad of combinations thereof. We aim to share our learnings of evaluating AI solutions in terms of quality, performance and cost trade-offs in the industry-leading MLPerf competition ("The Olympics of ML/AI benchmarking"). We describe an innovative approach based on automated reproducible workflows, which provide rigorous testing in development and continuous monitoring in production. Our approach is accelerator-vendor and server-vendor neutral, which helps de-risk hardware investments through best-fit selection and easy migration if necessary. The approach is applicable to a broad range of industries, in particular, heavily regulated ones: integrating checks and balances into automated workflows improves compliance and security.
Bob Oates, Associate Director, Cambridge Consultants Cross-domain approaches to GenAI assurance The deployment of an AI-enabled system into a critical system context involves multiple stakeholders from multiple domains, including security, safety, data science, software engineering, and AI development. How can MLOps processes be augmented to deliver systems that provide confidence to the user, the developers, and the accreditors of the system? In this talk the author presents a high-level assurance framework that attempts to reconcile risks from multiple domains and a taxonomy of mitigation techniques that work in an MLOps setting for a GenAI application. Important lessons learned about the opportunities and pitfalls of integrating assurance and risk management strategies into the MLOps process are shared. |
| 15:15 - 15:45 |
Break | Exhibition & Networking |
| 15:45 - 17:25 |
What does the future of AI look like? Chair: Julia Gwilt, Partner, Appleyard Lees IP LLP This session speculates on AI's trajectory, covering new models, hybrid systems, sustainability in AI training, multi-agent systems, and explainable AI, aiming to foster discussion on the future of AI technologies. Where will this technology be in 5 years’ time?
Peter Whale, Founder & CEO, Vision Formers Future AI technologies In this bite-sized session, we will provide a sneak peek at some of the emerging areas in AI technologies, and provide pointers to services and models that you can experiment with yourself.
Dr Jade Hind, Senior Consultant, Space, Defence and Intelligence, CGI Personalisation 2.0: AI's Role Ahead This talk delves into the transformative potential of Generative AI to create deeply personalised experiences. By moving beyond generic recommendations, AI can deliver services that truly reflect individual preferences, benefiting both consumers and businesses. It underscores the need for skilled professionals to design secure and adaptive systems, and the importance of balancing innovation with ethical and sustainable practices. Emphasising inclusivity and fairness, the talk addresses the environmental and economic challenges of AI. Ultimately, it envisions a future where AI provides highly customised and responsible solutions that significantly enrich our lives.
David Pollington, Head of Research, Bloc Ventures Where next? AI's evolution in knowledge, understanding, reasoning and scaling AI innovation is continuing at breakneck speed with advances being published weekly by both academia and the private sector. This session will explore a number of these areas including: improving AI knowledge and reasoning through agentic workflows; boosting AI’s understanding through causal and neuro-symbolic AI; and the importance of integrated photonics and neuromorphic computing in combating the compute scaling barriers.
Dr Detlef Nauck, Head of AI Research, BT The Future of GenAI There is more to Generative AI or GenAI than ChatGPT. While the public discussion and the media are dominated by large language models, and image or video creation, there is another development happening in the background that is bound to change how we do science and engineering in the future. The AI industry has learned from playing games and is applying it in drug discovery, optimisation and creating new materials. I'll give an outlook how the public and not-so-public developments are likely to impact us in the future. |
| 17:25 - 17:30 |
Closing remarks |
| 17:30 | Event close |
*Draft agenda, subject to change ahead of the event
