The School of Informatics: Introduction



the School of Informatics was created in 1998 through the merger of the department of artificial intelligence center for cognitive science department of computer science artificial intelligence applications Institute and human communication Research Center it is a world leading center of research and teaching in computation information and cognition it is located in the informatics forum and the nearby Appleton tower it hosts a cohort of over 1,600 students from over 75 countries worldwide it employs over a hundred and fifteen academics and 130 research staff research within the school is carried out across six Institute's each of these manages a portfolio of funding and specialist facilities academic staff are appointed to the school and choose to belong to one or more Institute's informatics is perhaps the word that people haven't heard before they are more used to artificial intelligence or computer science but we're called informatics because we look at the whole science of processing information be that in natural systems or artificial systems so it includes computers but also brains and so we're a very broad school with people doing everything from experiments with eye tracking and seeing how people process information by reading it from screens through to people designing new chips to build novel computer architectures our Institute is an interesting one to a large extent because it stands in a way at the crossroad of many other Institute's because we're quite interested in theory and practice and it's about making changes to the real world and being acting application is healthcare having applications in business we cover the breadth of spectrum from you know quite formal quite kind of abstract to quite applied there is a lot of connections to many different things that are happening right now on the one hand there is the data driven innovation ideas that I've filled the market these days and and there's a lot of questions about how do you use data responsibly how can you trust the people that use your data and that kind of thing it's also connected to formal verification which is a theoretical approach to try and build trustworthy systems that are correct you can hopefully get a system that can actually give some way of explaining what it did so you can understand how it achieves a particular classification of something or I tells you this is better than this for instance when it comes to looking at you know two images or making a decision in healthcare or in any particular kind of especially safety critical I think that's quite important and also I think from a more personal point of view for me I think it's not just about its foundation but it's also about trust and safety whether you can trust a particular algorithms so we encompass some very mathematical and theoretical work in machine learning we also do quite a lot of applications of machine learning we have the the schools sort of focus on computational neuroscience which is people who are trying to understand how the brain works and how it processes information from a computational perspective we also have the faculty who are really interested in other applications in biology so bioinformatics kind of work one of the major breakthroughs that that's coming quickly in my area is a global understanding of some of the basic biological processes we've had whole genomes that's been there for a while last last 20 years we've been getting genomes from thousands of species but we're now starting to get things like whole connectomes so the central nervous systems just know but how every single connection in a brain communicates from from one cell to another and what the logical flow of information out is with you that within that structure the big challenge for informatics at the moment is a huge resurgence and interest in artificial intelligence so at the center of that for a lot of what we do is how artificial intelligence can help people in the health disciplines so there are a lot of initiatives now to try and fund biomedical or artificial intelligence this is about using information it's within the NHS system and in experimental research laboratories to have impacts for people in their medical environment with their GPS their consultants their surgeons and we believe that there's a lot of information held in Scotland's unique medical record system which is the envy of much of the world because we have a very efficient and well routed system for linking individual patients to all of the tests and results and information that they have but there are also huge ethical implications about how we use artificial intelligence in the personalized medicine and so this school is also doing lots of work to try to understand how the impact of using these kinds of technologies has on people's rights and freedoms but also the benefits that it can have for helping their treatment of their condition in five to ten years I would say we will understand much better what the place of neural networks is in technological applications and and it will be a tool like other tools to use the future for the Institute will will continue to be well one aspect of it will be development of machine learning and artificial intelligence technologies the underlying technologies that can be applied in many different domains I think there'll be an increased aspect that we will need to pay attention to and we will need to do with the ethics and societal implications of these of these technologies the area that Wix Excel works in can be thought of as a as a bridge between the the requirements and desires of application designers what they can envisage might be feasible and the the other end on the other side of the bridge we've got the the raw communications technology the raw processing technology and we have to bridge that gap and what makes that interesting and exciting is that both sides of the bridge are changing over time the metrics the measurements we might make of what's a good designs are also changing so for example traditionally may be the dominant requirement was for things to be fast but nowadays it's equally impressive more important to make systems which are our energy efficient make good use of energy for example if you've got battery life to consider at one end of the spectrum you don't want your devices to die on you at the end of the spectrum you've got the energy requirements of a data center the main problem here is that given a new domain how do you design the computer for this new domain and that turns out that manually it's extremely costly right so what we are thinking now is to kind of automate this process of designing the next computer for the new domain and for that we are using these pillars of computer science right so formal methods and data science for this so that's what is most exciting to me I think that you know one of the really exciting opportunities is to be able to apply these machine learning advances to really rethink computer systems and again this is not just about hardware it's really rethink about the way we program computers the way we interact with computers and the way we build computers by integrating machine learning you know pervasively into every aspect of computer systems the future of computing is bright in in the coming years we will see we are surrounded by a lot of intelligent applications in our day to day life beats self-driving cars or robotics manufacturing health care and many more and in these applications we want to ensure that we can trust these applications because in the end there will be integral part of our life and ensuring safety and security of these applications is is something would be the grand challenge in the coming years Institute of reception action and behavior is computational informatics that relates to things that have to interact with the real world so this is either robots who are having to interact with the real world computer vision where you're trying to interpret information coming in from the real world or computer graphics where you're trying to generate information that then looks like the real world on the other side you may have heard recently that AI can do all kinds of wonderful things like recognize objects Drive cars play games but this takes a lot of data to happen you have to collect thousands of hours of driving data a thousands of photos to learn to recognize objects so we work on how to teach a is to do things with only one or two experiences much like humans can so they can learn more quickly I think that our field is at a cusp of actually being able to have a transformative effect in the external world and so we are finally at a point when we no longer have to explain to people what we do especially not official intelligence everybody knows it exists and a lot of what people are using might be based on technologies that people here have worked on so I think the next big step for us would be to go from small results that somehow creeped into products to much more ambitious kind of engagement you know entire companies coming out of the department you know our technology being at the forefront of big products we are heading towards smarter and smarter a I'm making more differences in people's lives whether it's going to be self-driving cars or better personal assistance helping you manage your daily tasks and everyday and informatics we are working on the underlying technologies which make those things come true Alexia's blab that I'm the current director of has quite a large number of faculty and it's something close to 40 faculty in it at the moment and they do a very broad range of research in a variety of areas of computer science including everything from research on programming languages research on databases research on algorithms and computational complexity research in security and cryptography research kind of modeling concurrency and probabilistic systems many many different sub areas the schoolb informatics is currently building up capability and skill into really primary areas which I bridge one of those men is human-computer interaction so how people interact with secure technology of any type and the other one is computer security both of which we've got a lot of new staff in and recently and I bridge those two spaces and so it's actually really fun sometimes I wander down to the Edinburgh School of Art sometimes I wander down and talk to the cryptographers and having both of those two communities kind of moving forward I think is really strong for the university there are problems whose inherent complexity is such that if you really want to build a specialized computer that solve the problem such computers will fill all the visible universe right to answer the problem in the reasonable amount of time so that's you obviously cannot do that all right so you have to go to the sort of notion of approximation and that's actually what the fields of data management and machine learning start converging so there's some very exciting work happening along these lines the future is hard to predict there's kind of a broad realization that many of the technologies that have been developed through in computer science over the past decades you know it's clearly been a success in terms of with the main technologies but I think with in more recent past we realized that those technologies have both a benefit and and a potential cost to society and and a lot of research in the future I think we'll probably be aimed at trying to mitigate the costs to society of the technology that we ourselves develop we need some way of kind of making sure that it's this future that is kind of unpredictable does not get out of control that we still have enough security and safety and and some amount of privacy like think of all the data that you're giving out at the stand process by algorithms you want to kind of know what is what is happening with that data and that it's not abused and that the decisions that are made about you a fair and unbiased the Institute is probably concerned three areas one of them is natural language processing so we have a large group that's working on things like machine translation there's a lot of speech work as well we're working on semantics on parsing cognitive modeling of human language processing another area is human-computer interaction and things like visualization some usability work educational technology as well and the third area that we're interested in is cognitive modeling where we model again language processing in humans but also things like decision-making language acquisition higher-level cognition including reasoning and inference generalization in humans natural language processing so that's dealing mainly with text trying to get the meaning out of text we're trying to turn go the next step and actually say that text out loud so we do a bit of things with people who do not rely image processing so you might imagine taking machine translation translating some text and then reading it out in another language you might imagine joining it up to speech recognition so then from speech to text to text in another language to speech so that's speech to speech translation that's kind of a hard really hard problem machine translation is an area of artificial intelligence so the way it works is that we teach software we write software that learns to translate from examples instead of looking at machine translations and saying this is terrible now people are actually struggling to find mistakes you know in some sort of easier kind of language easier sentences are often very well translated what is going to be the next step probably is things like automatic summarization and question answering at a more advanced level so question answering right now works at the level where you can ask for factoids how high is this mountains how many people live in this or that town but questions that involve some sort of reasoning and making inferences are still too difficult for systems to process and I think we're going to get closer to that over time well natural language processing I think there's two big goals we really want to be able to talk to computers have conversations with them and interact with them the same way we would do with other humans because that's for us the most natural interface of communicating and having to learn to use map and mouse or keyboard use these interfaces that's just a hassle for people that's a really big challenge because there's a lot of understanding that the computer needs to be able to do so that we can just talk to it and this will keep us busy for a long time the research profile of the informatics in Edinburgh squat in the last few years just as it has actually across the world but in many ways Edinburgh has been leading the way so and the reason for that is because historically Edinburgh hasn't just been a place where you can do computer science we were one of the pioneers in AI and in cognitive science and what that was about was bringing computational models and formal thinking to problems in the humanities that on the face of it looks like a mess without any regular structures or anything we believe that's not quite true that there are regular structures in these things and of course that's become the accepted view in informatics across the world in the future I think machine learning is going to continue to dominate more recently machine learning has become dominated by what's known as deep learning that's the application of neural nets to doing solving certain tasks but I think as things go forward that is going to change and we are going to analyze those models and produce more interpretable models it won't just be a black box anymore we will be mixing it up with things that we learnt before neuron models became dominant the approaches will become more hybrid and come together again I feel there are a wealth of opportunity for future research and informatics computers and computer decisions and artificial intelligence of becoming more more pervasive in people's daily lives their washing machine now has a computer that is as strong as what would have been on a researchers desk 20 years ago and so there are many many questions still to be answered as we find new uses and also as we do research we generate more applications and so it's a very exciting time to be a researcher in informatics

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *