What a PhD Is, Country by Country
What a PhD in AI involves, how the Australian, UK, European, US and Canadian models differ, what panels look for, and whether a PhD is the right step for you.
What a PhD in AI is (and is not)
A PhD trains you to do independent research: by the end you can find important questions, answer them rigorously and communicate the answers. In AI, the output is usually a series of papers (often 3–5) joined into a thesis. You are a junior colleague, not a student taking courses: progress depends on self-direction, persistence and your relationship with your supervisor.
A PhD is required for most academic careers and many industry research scientist roles. It is not required for most ML engineering jobs. It is a good choice if you enjoy open questions, can tolerate uncertainty and slow progress, and want to spend years on one area.
Models by country
| Region | Length | Entry | How funding usually works |
|---|---|---|---|
| Australia | 3–4 years, research only | a degree with a substantial research component (Honours, research Master's, or a coursework Master's with a research project; publications help) | competitive scholarship rounds (RTP and university awards): stipend plus tuition fee offset; a supervisor must support your application |
| UK | 3–4 years | Master's or strong Bachelor's | studentships through doctoral training programmes (UKRI-funded centres) or advertised projects |
| Continental Europe (Germany, Netherlands, Switzerland, Nordics…) | 3–5 years | Master's usually required | often a salaried employment contract on an advertised position; programmes such as ELLIS and IMPRS recruit centrally |
| USA | 5–6 years incl. coursework | Bachelor's or Master's | admission to a department with funding (research or teaching assistantship plus tuition waiver) usually guaranteed for several years; you choose an adviser during or after the first year at some schools |
| Canada | 4–5 years | often a research Master's first | funding packages from the supervisor's grants and university; institutes such as Mila and Vector recruit through member universities |
These are typical patterns; each university and programme sets its own rules.
What panels and supervisors look for
- Research potential: evidence you can do research (a paper, preprint, thesis or substantial project with code), and how you talk about it.
- Fit: your interests match the supervisor's and the group's direction.
- Strong references: specific letters from people who supervised your research.
- Academic record: grades, especially in relevant subjects (scholarship rankings often weight this heavily, together with publications).
- Communication and motivation: a clear statement or proposal; a good interview.
Many scholarship panels score research outputs explicitly (e.g. publications by venue ranking, first authorship). That is why Parts 1–4 of this course focus on producing a real paper.
Quiz
1. In Australia, a PhD scholarship application usually needs…
2. In much of continental Europe, PhD students are often…
3. The most persuasive evidence of research potential is usually…
📚 Further reading
Trusted, mostly free sources: official documentation, well-known textbooks and courses. When you want to confirm or go deeper on anything here, start with these.
- A. Karpathy, "A Survival Guide to a PhD" ↗ choosing problems, writing, and PhD life
- P. Guo, The PhD Grind ↗ a free memoir of a PhD