Meet Neil and Claire, our hosts for the day, who will introduce you to the world of AI (Artificial Intelligence) with experts from BT and King's College London:
2 - AI with Detlef Nauck
Neil (host): Now that we know what we're going to be exploring, let me introduce you to our first industry expert, Detlef Nauck. Detlef will be exploring how robots and autonomous cars learn to see and how this ability can be used to clean up the planet too.
Over to you Detlef...
Head of AI & Data Science Research at BT
Key qualifications: MSc, PhD, and a postdoctoral degree all in Computer Science with a focus on Data Science and Machine Learning
What does your job involve? The best thing about my job is that I can learn something new every day. I get to work with BT's incredible scientists to develop new and amazing AI technology and I work with colleagues across BT to help them make use of what we develop in BT's Research Labs.
How did you get into your current role? I joined BT Research over 20 years ago after 10 years as a University Lecturer in Germany. I have always been working in Research and last year took responsibility of running BT's AI research programme.
What did you want to be when you were younger? After briefly thinking I might go into medicine I came across computers at the age of 16 and that changed everything.
What do you do outside work? I love cinema and in particular SciFi and Marvel movies. In the real world I enjoy going for countryside walks and hitting the gym.
Detlef: Thanks for watching my video, I bet you didn't realise that you could use AI to sort cucumbers into different groups! We have had some questions submitted relating to this topic in the lead up to Norwich Science Festival which are answered below. However, if you have any questions, we'd love to hear from you. Please email email@example.com
Questions and answers
How does AI interpret images? How do they read and understand them?
To AI, an image is simply a matrix of numbers, it does not actually understand what the image displays. An image consists of rows and columns of pixels and each pixel has a colour value which is typically represented by three bytes (or 24 bits) – one byte each for the intensities of red, green and blue in the pixel. A small image of 100 by 100 pixels would already contain 10,000 pixels that are represented as 30,000 bytes (or 240,000 bits, each being a 1 or 0). To create a more efficient representation that isn't quite so big, the pixels are typically turned into grey-scale images. Each pixel is therefore represented by a single number between 0 and 255, which is the range one byte (8 bits) can represent. An AI system for image analysis – a so called 'convolutional neural network' – then looks for patterns in the image to classify it. For more information on this, take a look at the following site: Convolution Neural Network for Image Processing.
Could AI take over humanity?
Although this is a frequent topic of Hollywood movies, it is actually highly unlikely that AI could take over humanity. This is a common myth amongst many other AI myths and you can read more about them here: The Top Myths About Advanced AI. Although AI can go wrong and cause problems if it is not built/controlled properly, for it to take over humanity it would first have to become much smarter. This is also called "artificial general intelligence" or AGI, and means that an AI system has human-level intelligence. Current AI is far from that and only qualifies as "narrow AI", which means that the AI system can only focus on one very specific problem e.g. recognise images or answer specific questions. Nobody so far has made any real progress towards AGI and most researchers agree that we are decades away from it. That doesn't mean we shouldn't be prepared and make sure that all AI we produce is benevolent and aligned with our values.
Claire (host): Now it's your turn to have a go. You're going to teach an AI system how to recognise cats as well as things you would expect to find in a pencil case. Can it then spot the difference between the two? You can download the activity pack to get step by step instructions.
Neil (host): Now that you know a little more about AI, take some time to explore more roles in this space by taking a look at these profile cards...
Artificial Intelligence (AI) and Machine Learning (ML) Solutions Architect, BT
Key qualifications: Masters in Maths. Experience working in ML and Google.
What does your job involve? I work across the business building the platforms on which our data scientists build and run their ML and AI algorithms. I also work with lots of external companies to understand what new technologies are out there and which will help BT grow.
How did you get into your current role? I started as a graduate researcher in BT's Data Science team where I completed a number of projects before moving around the business doing a variety of data science roles. From there I decided that instead of building the models I wanted to shape and develop the actual technology and inform BT's strategy around the future of its AI and ML offerings.
What did you want to be when you were younger? Professor of Mathematics
What do you do outside work? Lots of sport! Rock climbing, badminton, tennis, surfing and skiing to name a few. I'm also grade 8 on French Horn and Piano.
5 - King's College London
Claire (host): Meet Michael Luck from King's College London, who will explain how autonomous systems can work together to perform more effectively and efficiently.
Over to you Michael...
Professor of Computer Science at King's College London
Key qualifications: PhD in Computer Science
What does your job involve? At King's College London I work on artificial intelligence and in particular on the AI that's required to get computers to work together effectively, much like people do. I also lead a programme to train PhD students in methods of safe and trusted AI.
How did you get into your current role? As an undergraduate, my final-year project involved some research into AI for scientific discovery, and I continued that when I undertook graduate study in the US. That led me to come back to London to do a PhD in the same area, and to ask questions about not just about the process of scientific discovery by machines but also about the goals or intentions of the system (human or machine) doing it. I've been working on different aspects of AI ever since.
What did you want to be when you were younger? Writer or architect
What do you do outside work? When I've been able to do so, I've recently enjoyed going with my father to see our football team, Chelsea. It's been a brilliant way to spend time together.
Michael: Thanks for watching my video. I hope it's given you a great insight into what's involved in intelligent machines collaborating with each other and with people as well as ensuring that we can trust them. We have had some questions submitted relating to this topic in the lead up to Norwich Science Festival which are answered below. However, if you have any questions, we'd love to hear from you. Please email firstname.lastname@example.org or join us live between 13:45 - 14:30 on 12th October for the Q&A session.
Questions and answers
Where do you think AI will help humans the most in future? Which type of jobs could it assist with most?
In principle there is no limit to what AI might be able to help us with in the future, but in the immediate term AI is more likely to help us with predictable and repetitive tasks in a controlled environment. That is to say that AI works best when we can anticipate what is likely to happen. So the types of jobs are those in which there's a lot of drudgery: trawling through legal cases for key information, routine administration and various manual but well-defined tasks. AI is much less likely to be able to help us in jobs where there is a lot of creativity or where the environment is very uncertain.
What can be done to help people build trust in AI?
The most valuable thing to help build trust in AI is to be able to explain how an AI system has come to a particular decision. For example, if you've applied for a loan and the computer says no, then you need to be able to understand exactly why that decision has been reached, and to be able to trace through the steps and the information to make sure it's understandable and correct.