MSc Artificial Intelligence

The universities worth studying AI at are the ones that were doing it before it was fashionable — places with decades-old research groups in machine learning, natural language processing, computer vision and robotics. Their master's programmes teach the mathematical machinery underneath modern systems rather than tool use alone, because the tools change yearly and the mathematics does not.

Admission has become intensely competitive as the field has boomed. A credible application shows mathematical maturity and some evidence of hands-on work — a project, a thesis, a contribution — rather than enthusiasm for the technology's headlines.

A demanding specialist master's that assumes both the mathematics and the programming are already in place.

• A bachelor's degree in computer science, mathematics, engineering or physics.
• Linear algebra, calculus and probability at undergraduate level — non-negotiable, because the core methods are built from them.
• Confident programming, usually Python. This is not a programme that teaches you to code.
• English proficiency evidence.
• A statement engaging with a specific area — vision, language, reinforcement learning, robotics — rather than artificial intelligence in general.

This is the most oversubscribed field in computing at the moment, which makes admission competitive and makes the mathematics prerequisites more strictly applied than they once were.

Genuine demand, concentrated in fewer employers than the coverage suggests, and increasingly split between building models and applying them.

• Machine learning engineer — the largest destination, and largely software engineering with a modelling component.
• Applied scientist and research engineer roles, generally at larger organisations.
• Computer vision and NLP specialist positions.
• Robotics and autonomous systems, especially where the degree included control.
• Doctoral study, the normal route into genuine research.
• Domain applications — medical imaging, drug discovery, climate modelling — where subject knowledge plus these methods is rare and valuable.

Research positions at the frontier generally require a doctorate. A master's leads to applying and engineering these systems, which is where the great majority of the actual work is.

One year, technical throughout, moving from the mathematical foundations to a chosen application area.

• Machine learning foundations — the theory, not just the libraries.
• Deep learning — architectures, training and the practical difficulty of making any of it work reliably.
• Natural language processing, including the transformer architecture underlying current language models.
• Computer vision and, on many programmes, reinforcement learning and robotics.
• Ethics, safety and interpretability, now core rather than an afterthought and increasingly what employers ask about.
• A dissertation, often research-oriented and a common route into doctoral study.

The field moves faster than syllabuses do. What the year should give you is the mathematical foundation to follow it afterwards — the specific tools taught will have changed within a few years.