• Richard Harvey

How to study AI?

A had a kind invitation to speak at Uni Taster Days. The idea is that potential students tune in on a Tuesday and get some information on studying that particular subject at university. The nub of the issue however is, is it worth studying AI at university or should you study something more general such as Computer Science or, even, whisper this, Physics or Chemistry!

The first point I would make is incredible conservatism of employers is hard to under-estimate -- there is a story, maybe apocryphal, of a briefing from the Head of HR at the BBC who claimed that the Beeb had never, knowingly, employed a Media Studies graduate. Accordingly I rang a couple of friends who work in Data Science and/or AI companies. Predictably they were incredibly conservative and mentioned a wide range of numerate degrees as providing suitable backgrounds for their excellent companies. When I pushed them a bit a number of themes emerged:

Firstly, their own backgrounds were quite diverse so, naturally they tended to assume that anyone could follow their own learning path. In my experience that is not true but its nice to hear that employers are optimistic.

Secondly, it is not good economic sense to limit entry to competitive jobs too narrowly. If one only accepts applicants with BScs in Underwater Basket Weaving then clearly there will not be many applicants.

Thirdly, programming is an important skill in AI. This is a major challenge for employers. There is simply no guarantee that a candidate who studied Physics or Chemistry can program in any competent manner. And they might not be able to learn (or rather learn enough to be good enough).

Fourthly, AI is important but knowledge of other things such as statistics, computer architecture, programming and, nowadays, ethics and professionalism are all critical. And even more important is what we call "domain knowledge" which is understanding the domain to which AI will be applied.

So the message is clear. As far as employers are concerned, there is no pressure for universities to run degrees in AI. But, that is not the same as saying that studying AI is valueless. Far from it. Nor is it the same as saying that the traditional view which used to be, if you studied Maths or Physics then you were ready for anything is true either.

But this is all very confusing for university applicants you say...universities are putting on these innovative courses and employers don't want them. Yes and no. Yes in the sense it is completely illogical that employers should want to employ anyone who studied Latin and Ancient Greek. No in the sense that there are some enlightened employers out there and the tide is turning.

So, what really matters? I can illustrate this with a story - again I hope I have remembered this correctly but, for safety let's say it's apocryphal. A university where I worked decided to measure the time at which students in their first week of term handed-in their first piece of work. What they found was that there was zero correlation between the mark awarded for that work and their final mark after three years of study. But, when the took account of the time, they found that people who handed-in the work first were those who got First Class degrees. What does this story tell us? I infer that motivation and enthusiasm are absolutely critical to a successful degree outcome. If you want to get a good degree then choose a course that you find interesting. And that's the secret to choosing a good degree, whether in AI or something else.

14 views0 comments
  • Twitter

©2020 by Richard Harvey, Professor of Computer Science and AI, Public speaker and consultant.