This article is part of a series of articles on different European master's programs related to artificial intelligence and machine learning.

Basic data on the degree

• Duration: 1 year (or 2 years part-time)
• Cost: ~15.000€ p.a. for EU students (before Brexit)
• 90 ECTS

Purpose of this article 

The main focus of this text is to help you decide, whether studying an MSc AI degree at the University of Edinburgh (AI@Ed) is an option for you. It is not supposed to give detailed technical information like links to admission pages or tips for how to find a room in Edinburgh. You will need this kind of information once you decide to apply to UoE. I hope that I can guide your decision-making process with this article. 

This is also not an advertisement, I have no affiliation with the University of Edinburgh. It's just an honest and objective opinion on the degree, the university, and the city.

Text is by: Marco Kinkel, feel free to message me on hi@marco-kinkel.de

The degree

Courses and Areas

Being able to choose from a variety of topics was important to me. AI research is manifold and includes grasping a notion of intelligence from different viewpoints. I was happy to choose from interdisciplinary courses on Cognitive Science, Neuroscience, Robotics, Bioinformatics and of course many core ML courses. The university is especially known for its research and lectures on Bayesian ML, Computer Vision, Natural Language Processing and Biomedical Sciences. Here is a list of most important courses (which obviously changes over the years). The bold courses can be considered the standard of the degree, you will find the majority of your friends in these. However, depending on your interests, it can be worth thinking outside the box and taking a less popular course. It can expand your horizon and you will get better teaching because they are less crowded (e.g. 10 instead of 200 students).

• ML
  • Introductory Applied Machine Learning
  • Machine Learning and Pattern Recognition
  • Machine Learning Practical
  • Reinforcement Learning
• Vision
  • Image and Vision Computing
• Language
  • Accelerated Natural Language Processing
  • Automatic Speech Recognition
  • Natural Language Understanding, Generation, and Machine Translation
• Data Science
  • Text Technologies for Data Science
  • Data Mining and Exploration
• Design and HCI
  • Case Studies in Design Informatics
  • Human-Computer-Interaction
  • The Human Factor: Working with Users
• Bio and Neuroscience
  • Computational Cognitive Neuroscience
  • Bioinformatics
• Misc
  • Probabilistic Modelling and Reasoning
  • Computational Cognitive Science
  • Natural Computing
  • Robotics: Science and Systems
  • Algorithmic Game Theory and its Applications
• Introductory Informatics Courses (for students from other fields)
  • Introduction to Practical Programming with Objects
  • Computer Programming for Speech and Language Processing
  • Programming Skills
• Additionally, you can choose one or two courses from other schools, including courses like
  • Robotics, AI and the Law
  • The Computational Mind
  • Ethics of Artificial Intelligence

You can find descriptions of these courses in the Degree Regulations and Programme of Study (DRPS) for the academic years 2019/2020 and 2020/2021. Unfortunately, the current Covid-19 situation results in a cancellation of many courses for the academic year starting in September 2020, as you can see in this list.

As you can see, AI@Ed does not force you in a particular area of AI. Many courses exist on Machine Learning and Language Processing, but you can always choose to flavour your degree with diverse (but due to the time limit of one year slightly superficial) knowledge in AI-related fields like robotics, neuroscience, language processing, philosophy of mind and computational psychology. The variety of courses has its drawbacks: it is difficult to choose only 6 to 8 courses within the very short time of one year. This is a general downside of the degree to which I will come back later. From an EA standpoint it is important to notice that while some courses mention AI ethics or AI sustainability, no course specifically focuses on AI alignment or safety. 

In addition to the choosable courses, you will have two mandatory courses that are supposed to prepare you for the final thesis: Informatics Research Review (IRR) introduces academic writing and citation techniques and is just additional practice in academic writing. Informatics Project Proposal (IPP) on the other hand is actually useful because you collect literature and write a proposal document for the topic of your dissertation.

For people with a less technical background, there are courses to enhance your programming skills. These are useful but not necessary because most courses don't require a lot of programming experience. My friends with backgrounds in Physics, Neuroscience or Statistics easily succeeded without these additional courses. If you have used a Jupyter Notebook before and used Pandas, R or Matlab to inspect some experimental data, you’re all set. You don’t need any knowledge on subject-specific Python libraries, as those will be introduced in the respective courses. A good introduction to ML-related Python libraries is the labs of the lecture Introductory Applied Machine Learning.

Generally, the amount of code you will encounter depends a lot on your courses. In most courses the lectures impart theoretical knowledge which is applied in their tutorials and labs. If you’re interested in coding and practicing ML model design, implementation and parameter tuning, you should take Machine Learning Practical. Other courses like Game Theory are purely theoretical.

Influential Researchers at UoE

• Chris Williams (Turing Fellow; ML, statistical pattern recognition, probabilistic graphical models and computer vision)
• Mirella Lapata (computational models for the representation, extraction, and generation of semantic information)
• Amos Storkey (Bayesian and Neural Systems)
• Sharon Goldwater (unsupervised learning of linguistic structure)
• Stephen Renals (development of interactive systems that can understand human communication)
• Iain Murray (Probabilistic Machine Learning and Inference)
• Peggy Series (Computational Psychiatry)
• Simon King (Speech processing)
• Michael Gutmann (Probabilistic Modelling and Reasoning)
• Matthias Hennig (Computational Neuroscience)

Lectures, Tutorials and Labs

You will have lectures with between 50-150 other students from different Informatics MSc degrees (Informatics, AI, Data Science, Cognitive Science, ...). Most main lectures are not very individual with 150+ students, but the tutorials and labs are divided into small groups. In tutorials, you typically discuss exercises that go somewhat deeper into the lecture material. The tutors are mostly Ph.D. students who can be more or less motivated or talented for a tutoring job. In labs, you complete assignments and other exercises with the help of (often not very helpful) instructors and your fellow students. The attendance to lectures and tutorials is only mandatory for non-EU students due to visa regulations (note that this might change after Brexit). 

The lectures have very different qualities. In the UK, I think, professors are not obligated to hold lectures, which increases their interest and hence the lecture quality. But in my personal experience, one of five lectures is still so bad that it may be better to just read the slides (but this is probably the case in every university).

You will encounter some 'inverted classroom' lectures, where the material is provided as videos and the actual lectures are QA sessions. There are very few courses where you do actual research (that could be published). In Machine Learning Practical where you develop and evaluate your own ML models, you have a lot of freedom, which comes closest to actual ML research.

Generally, all lectures are recorded, so you can re-watch them as often as you like. This is very useful in the coursework period (middle of semester, see below), where you will hardly have the time to attend all lectures. 

Most courses contain two to four hours of lecture plus one or two hours of tutorials and labs per week. A typical week in my first semester looked like this:

You can see that it is quite full if you attend all tutorials and labs (which you should in the beginning to keep up the pace). The material is conveyed very quickly, so being sick for a week is not a rewarding experience. Between lectures, you will find yourself sitting in Appleton Tower (see below) doing assignments and catch up on lecture materials. This leads to the second point of criticism: the work-life balance is rather bad. Since every lecture has mid-term assignments worth between 10 and 50% of the final mark, it's difficult to keep up with the lecture material during the semester. Hence, you will not have a lot of fun in the weeks of the exam period. I cannot compare the work-life balance in AI@Ed to other MSc degrees from my own experience but I've heard of degrees with less stressful semesters and exam preparations.

Exams and Dissertation

The UK has its own marking scheme, which seems pretty self-explanatory at first glance:

Everything below that is a fail. However, a mark higher than 75% is only given for extraordinary performance, i.e. if the quality of your work is publishable. Therefore, many exams contain open questions worth 25% to keep students from getting too high marks. The general range of grades is therefore somewhere between 60 and 75 percent. This can be confusing for recruiters outside of the UK, who might think that a 73% degree is rather bad. Note that the 'excellency' range is as large (25%) as the rest of the marking range, so whether your work is deemed excellent, and if so, how excellent, is rather unpredictable.  From our experience, excellency sometimes just means a huge amount of extra work, but this is again not guaranteed to give you >75%. This critique applies to all universities in the UK, not only UoE.

Usually, you will have one coursework per course within the semester. It consists of applied exercises on the lecture material, but it can go way beyond. Completing it will take a considerable amount of time, which is stressful since you have a coursework (CW) for every course. Your effort should depend on the weight of the CW, which is usually between 10% and 50% of the final course mark. However, the CW is marked by many different tutors in a more or less unmoderated process, so you can have frustratingly bad luck with your marker.

In contrast, the dissertation process is organized very well. You can propose a thesis project topic yourself or you can choose from a huge list of offered projects. The selection process is not interesting here but be assured that you will definitely find an exciting project in your field of interest, some even in collaboration with Amazon (although it's hard to get them). You will start with the dissertation project after all exams and lectures are completed (mid May) and finish after 3 months (mid August). This is a very short time for a dissertation, but most of the supervisors are aware of that and hence define the topics narrowly. I don't know about anyone who published a research article from their dissertation but this should be possible. However, bear in mind that you only have 3 months. This is a downside for people aiming for a PhD after their degree. But don’t despair,many MSc thesis supervisors will offer you a PhD position if you did well in your thesis.

Here is a public archive of outstanding dissertations from 2019, to get a glimpse of the variety.

One-year degree

Although it is compelling to do a one-year MSc degree, to quickly proceed to the next level, this short time also has its disadvantages:

• No contact to master students in higher year
• you have to think about the thesis (self-proposed) and a Ph.D. (application deadlines) very early on
• you can take only a few courses
• you don't have a lot of free time to explore the city and country, because the dissertation is written in the holidays

I want to address the problem of specialization again: If you already know which field to specialize in, you're fine with a one-year degree. If on the other hand, your aspiration is to get a broad interdisciplinary knowledge about AI, that's perfect too! But you cannot get both in one year. Say you spend your 6-8 courses on Introductory Applied Machine Learning, Computer Vision, Natural Language Processing, Computational Cognitive Science, Robotics, Computational Neuroscience, Data Mining, and Game Theory. Then you had one course in each of those fields, which is great for a broad overview but rather bad if you want to call yourself a professional afterward. I would advise against doing a one-year degree if you still want to explore the diverse AI-related research areas and specialize in some sub-field.

University and organization

Admission

As for most AI programs nowadays, it's very hard to get in. First of all, you need two reference letters and good grades. According to this website, the offer rate is about 14%. This coincides with my experience of having exceptionally smart and diligent fellow students.

Buildings, working areas

UoE's lecture halls are very modern and large. If you don't take unusual courses, they are all within a radius of a 5-minute walk and all within the south of the city. A nice park (the Meadows) is nearby where you can spend some time in the sun (if it's out).

As a busy student, you will spend most of your spare time studying in Appleton Tower, that is a modern 9-floor building dedicated to informatics students. It contains lecture halls, large lab spaces with PCs, and seminar rooms which can be used for studying in groups or individually. In exam times you can find people basically living there, which is possible thanks to the kitchen areas. When you see the modern interior equipment in the working areas, labs, and lecture halls, you know where the student fees go.

While Appleton Tower is specifically for School of Informatics students, the main library is open for everyone and offers additional workplaces. However, in the exam period, you will have difficulties finding a spot there.

Every big building like Appleton Tower, David Hume Tower, and the Library have a small cafe on the ground floor. Here you can get hot beverages, snacks, and at lunchtime even a small selection of hot food. I will address the food problem in a second.

Organisation, ITO, Student Reps

Officially, the university is very open for communication and there are many channels to approach. We have a very friendly staff at the Informatics Teaching Organisation (ITO), who is responsible for all official student-to-university issues (exams, lecture organization, tutorial group assignment, ...). The ITO together with the student representatives work very well for organizational issues. However, it is very difficult to get in touch with the lecturers, the researchers, and their departments. Bad decisions in the university's upper management led to a massive influx of students in the past years, which overwhelmed the staff and led to long chains of communication. The high-profile researchers are flooded with requests for supervision and have to cut their research to cope with organizational tasks. If you are really interested to work with one specific researcher, you will eventually get in contact with them but it's difficult to get a broad overview of the research at UoE because you can't just walk in the departments and have a chat. You cannot even get close to the research departments, because they are located in a building you can only enter with an appointment. This also frustrates the researchers at UoE, which is why they participate in strikes.

Strikes

During my one year in Edinburgh, we had two strikes organized by the University and Colleges Union (UCU). The reasons for the strike are manyfold. One of them is the massive increase in student numbers (which increases the UoE turnover) together with the decreasing spending for staff and organization. You can find more reasons here.

Many lecturers and academic employees participated and even students solidarised, which led to buildings being closed, and some lectures being canceled. The demands have not been fulfilled since, so additional strikes can be expected in the next few years.

Clubs and Activities

A very curious component of university life in the UK are clubs. You will find a club for every thinkable hobby or interest (Harry Potter Club, Skydiving Club, Atheism and Humanism Club, Beneficial Artificial Intelligence Club, etc). Clubs always welcome new members and are a great opportunity to try new activities. However, the program is so stressful that you don't have much time for activities anyway. Most people's free time activities were restricted to going to the gym. This of course depends on how ambitious you personally are, there are rumors of people who actually have time for other extracurricular activities.

Edinburgh

Food

The university, unfortunately, has no central canteen with cheap food. There is a large selection of wrap, soup, and sandwich places in the area and the cafés in the university buildings also sell some food, so you will not starve, but the lunch becomes rather expensive over time (e.g. £5 for a wrap. If you fancy a nice hummus falafel wrap I recommend the instagram account that writes in-depth reviews on each of them in Edinburgh). It makes sense to bring your own food and use the microwaves in the university buildings.

Accommodation

The university dorms are usually more expensive than private housing and you can only apply to them after you received an unconditional offer (which happens rather late), so I would recommend looking for private shared flats on SpareRoom. The rent is high in Edinburgh, you can expect to pay £450 to £650 per month for a room.

Edinburgh

Edinburgh is a beautiful city in a beautiful country. Although it has half a million inhabitants, it feels like a small town if you avoid the tourism spots. This is easy as a student because the university area is south of the tourism center. If you live close to the uni, you will not need any public transport ever. Uni, pubs, supermarkets are all within a walking distance.

The city also has a lot to offer, with many (over-priced) attractions such as the castle, but also beautiful and free places such as Holyrood Park. The city offers lots of pubs, nice places to eat, and cultural activities. You can easily avoid the tourism areas (except in August with the Fringe Festival).

The winters in Edinburgh are cold, windy, rainy and dark, and depending on your accommodation, going inside doesn't help much. So prepare for that by bringing warm clothes and buying vitamin D supplements and a SAD lamp or light box. However, the summer is beautiful and if you have the time you can swim in the sea, go hiking or just roam around in the green parks.

Summary

AI@UoE considers itself as an 'elite' program in Europe. Considering the acceptance rate and the intelligence and diligence of the students, this is definitely true. However, it depends on your course choices and a bit of luck, whether you receive an 'elite' education. Some courses and lecturers are more challenging than others, which often leads to a better learning outcome, but a worse work-life balance. I'm not sure where the pressure comes from, but some lecturers feel the urge to compete with other elite universities when it comes to the course contents and speed ("We cannot cut topics because we must compete with a similar course at the University of Oxford''). This can be frustrating, but again, you will learn a lot more in these challenging courses. 

Many high-profile researchers work at UoE and this degree can be a great starting point for a subsequent Ph.D. with one of them. However, the general direction of the university described in the strike section increasingly demotivates the staff and leads to less 1:1 communication for students, including the increasing difficulty to get in touch with those high-profile researchers. This will only become a problem if you are planning to build a network and connect with the local research staff. If you plan to just get your degree and move on, you will probably not be impacted by this issue.

AI@UoE offers a wide range of lectures with a good portion of most AI-related topics. You can get an interdisciplinary degree (including philosophy, neurobiology, psychology, cognitive science, robotics) or focus on core ML ideas. However, you will not find many courses covering EA related topics like AI Alignment and social impact of AI. The university is very modern and provides nice spaces to study and to collaborate. Finally, Edinburgh is the perfect mixture of a large city with lots of activities and a small town where you can live and study without being distracted by tourists. I would 

You should do the AI@UoE degree, if...

• ... you know which fields you want to specialize in
• ... you want to do a quick 1-year degree
• ... you like Edinburgh and Scotland

You should consider not doing AI@UoE, if...

• ... you want to gain broad knowledge about AI and a specialization in a sub-field
• ... you desire 1:1 communication with researchers
• ... you want unimpeded progress without possibly being affected by strikes
• ... you are more interested in the societal impact or possible beneficial applications of AI than in technical aspects