Focus July 2020

Principal AI Lead, BMT Esin has significant experience in AI for the last 30 years. She has developed several AI based solutions in Computer Vision. She has delivered many projects using earlier AI techniques to recent Deep Learning algorithms for Defence and private industry. She has been PhD supervisor to several students researching AI in the fields of Computer Vision and Knowledge Representation at leading UK Universities. Dr Turkbeyler now oversees a team of Data Scientists and AI specialists who apply AI in a huge range of fields, from satellite data, imagery, text data, natural language processing to cybersecurity. Esin Turkbeyler “It also explains why human-machine teaming has thrived, because at what point do we let a computer make the decisions and what happens if an algorithm gets it wrong?” After earning her PhD, Dr Turkbeyler started applying AI theories to solve real- world problems. She worked in the field of computer vision, processing video footage by recognising images and detecting and tracking the movement of people. “The development of computer vision learning algorithms was a great field to be working in at the time”, she says, “Humans take for granted the speed at which our brains can process the complexity of what our eyes are seeing, but computers struggled to make sense of these images, detect movement, and identify a person between different field of views. Identifying problematic scenarios “The quality of video footage from static surveillance cameras can vary, as can the lighting, so we’d need to identify each problematic scenario and programme an algorithm to help a computer understand how to deal with it. “With footage from a moving drone, for example, a computer would perceive everything to be moving, including the ground and buildings, so we’d have to programme algorithms that could differentiate between a building that’s actually static and the movement of people.” Data fusion and decision-making Towards the end of the 2000’s, Dr Turkbeyler became Technology and Team Lead in Data and Information Fusion, where she and her team integrated data from multiple sources and sensor outputs such as sound, images and radar to enable computers to create a complete operational picture. Their work with Bayesian networks, taught algorithms not just to distinguish images but to start making decisions for themselves. This is the same technology that’s used in autonomous ships and vehicles. “This was towards the end of the 90’s”, she says, “when several scientific theories and techniques converged into AI.” But what exactly is AI? “Well,” says Dr Turkbeyler, “many of our customers ask the same thing and really, they just want to understand how it can benefit their businesses. Artificial Intelligence algorithms”, she says, “are learning algorithms. “ “They aren’t rules-based. They reason. Some even self correct, learning from successes and failures. In short, they use computer systems to simulate how human intelligence works. “AI can help make complex decisions at great speeds based on vast quantities of data, rather than human interpretation or instinct, which can be slow or flawed. They detect patterns, anomalies and threats and can produce predictive analyses to forecast what’s most likely to happen next and, therefore, what the best available option is likely to be. This is why they are so good for predictive maintenance and cyber security as well as scheduling and planning. “Given the amount of data that organisations now have access to, you can see why they want to understand it to make accurate analysis and predictions. And given its potential to save time and money, it’s clear why data has become so valuable. It was Dr Turkbeyler’s natural interest in maths and physics that guided her towards a degree in electrical engineering, but her journey into Artificial Intelligence really began after she won a NATO Science Scholarship and started her PhD at Imperial College, London. The Role of Artificial Intelligence in Digital Transformation A conversation with BMT’s Principal AI Lead, Dr Esin Turkbeyler [email protected] 21 14 20 Ask Esin Turkbeyler, BMT At what point do we let a computer make the decisions and what happens if an algorithm gets it wrong? “ “ Cont’d