Graduate Series: An Interview with Heriot-Watt student Angus Addlesee
Hi Angus, firstly could you tell us a little about yourself and your area of study?
Thank you for inviting me to chat about AI. I am currently studying an MSc in Artificial Intelligence and prior to this I completed a BSc in Mathematics. During various internships I have created optimisation algorithms for Tipico, the leading sports betting company in the German market, and worked on data analysis with a financial syndication platform in London.
Register to get started
What drew you towards studying Artificial Intelligence?
During my time at Tipico I created purely mathematical algorithms as that was my area of expertise. At that point I realised that machine learning could enhance my programs. It wasn’t until I had graduated and started working as a data analyst that I fully realised the potential for AI. My job involved looking over lots of investment data and I noticed that the number of investments into AI related companies was growing all across Europe. I finally decided that I had to get involved and applied to study an MSc in Artificial Intelligence.
Your specific area of study is in multimodal and speech interaction with AI, do you have any insights on this application of AI?
My Mathematics dissertation focused on exploring Markov Models which was a great basis to understand the current methods used by AI to understand human speech. Current speech recognition software relies heavily upon Hidden Hierarchical Markov Models to imitate how the human brain recognises speech. I highly recommend “How to Create a Mind” by Ray Kurzweil to anyone that wants to read a thought-provoking and understandable book on creating an artificial mind.
Speech recognition is widely used today in systems like Siri and Amazon Echo but we will use it for almost everything in the future. I am currently working on a multimodal system to help the elderly in which speech recognition is essential. As the population ages we will need AI systems to assist our already overworked carers and nurses. These systems can detect falls, ensure medication is taken, order groceries and much more. I think that speech recognition is essential as it is important that elderly people at home can call for help in any situation.
Your approach to AI is both mathematical but with a strong emphasis on coding, how is your study time divided and how important is it to grasp both of these areas when studying AI?
Programming is extremely important when studying AI. R and Python are the two most important languages to learn in my opinion and I highly recommend learning SQL. You can study AI without too much mathematical knowledge but the most cutting edge AI problems are highly mathematical. Improving driverless cars and speech recognition relies heavily upon mathematics for example. Most AI companies will hire both computer scientists and mathematicians but I do feel that it is beneficial to have studied both. The majority of my time is spent learning the theory behind AI and the rest of my time is spent coding and completing projects.
Artificial Intelligence is at a particularly exciting stage and is increasingly used to optimise an array of different industries, from finance to healthcare. Do you feel Scottish Universities are promoting and encouraging this growth at the student level?
Honestly, I feel that Scottish Universities could do a lot more to promote Artificial Intelligence courses as it is not an area that school students initially think of when considering their future. Scotland is at the forefront of innovation in robotics and artificial intelligence so it would be great to encourage kids to view robotics and AI as realistic career paths rather than as science fiction. Even with a great interest in artificial intelligence I had to actively search whether courses existed, so more promotion is definitely required.
Where do you see the future of AI progressing and its wider societal impact? Is it a dystopian future akin to Terminator, or is it a little more nuanced with a focus on augmenting existing processes within our society?
Creating an AI with human level intelligence is an incredibly difficult challenge. Creating a safe AI with human level intelligence is an additional challenge on top of that. There is a risk that the first challenge is tackled before the second if we are not careful. Another issue to consider is that electronic circuits function much faster than biological ones, so an AI of human intelligence would complete 20,000 years of human-level intellectual work in just one week. Humans will continue to create more intelligent machines so at some point, Ray Kurzweil predicts 2045, humans will no longer be the most intelligent lifeform on the planet. This sounds apocalyptic but the leading researchers have already started work on methods to ensure that this superintelligent AI assists humans, as intended, rather than disregard us.
Sponsor a graduate article
To Find out More:
e-mail us on firstname.lastname@example.org
call us on 0845 643 5375
"Graduate Series: An Interview with Heriot-Watt student Angus Addlesee" first published on Company Connecting February 2017