More and more CS students are interested in AI – and there aren't enough lecturers Computer-science departments across US universities do not have enough lecturers to teach increasing numbers of students interested in AI, a report from the Center for Security and Emerging Technology (CSET) this month suggested. Interest in machine learning and artificial intelligence has risen and fallen since the field was …
But I wonder whether they're getting sufficient grounding in the first principles of computer science that will enable them to go on to creating robust, secure systems by learning to think in a flexible way about the subject. There's an increasing tendency in computer science education over the last few decades to gloss over the common groundwork in favour of concentrating on the current flavour of the month. This inevitably leads to scrappy understanding bolstered by assumptions, leaving problems imperfectly solved. We only have to examine the ease with which computer vision systems can be fooled to recognise this - and ask the question whether there might be an alternative way to automatically analyse images that is less easily fooled. Instead of which we have anchored ourselves firmly to one basic approach (that is fragile) and spend huge efforts tweaking it to 'improve' what may be fundamentally flawed.
Proper computer science is just a specialised branch of maths :-)
The Donald showed us that.
Do students still study things like Taylor series and numerical and operational analysis, or have we consigned them to ‘just include this library’ ?
However, maths is not enough if you actually want to build a computer as engineering ≠ maths.
Indeed virtually all of the summer and placement students we get say this.
They have no idea what they want their AI or ML to achieve, but they all want to do it.
They are somewhat disappointed when we say that we don’t do this AI thing as it doesn’t really provide us any benefit; there seems to be this belief that every problem and every business needs AI mumble.
But generally at the end of their 3 / 6 / 12 months they have learned a load of stuff that universities don’t teach. Like how to work in a team, what source code control, automated testing and release management means, why stuff doesn’t get pushed to production immediately, why we have several environments and so on … real world stuff.
It shows them things they never imagined and that they would never learn at university.
And that there are loads more things that are just as interesting as they believe their AI to be …
Generally they are much better for it.
But arguably that’s what placements should do and if we help somebody on their way then that’s a job well done.
>They have no idea what they want their AI or ML to achieve, but they all want to do it.
It avoids trying to understand the problem and possible algorithms to solve it.
In computer vision I have seen "AI" solutions to basically fitting a straight line to a bunch of points. If your onyl tool is TensorFlow - everything looks like a nail
Undergraduate degree programs are very limited in the amount of time they can devote to the major, at least in the US. There are general-education and breadth requirements which, however necessary or valuable, occupy many of the course slots over the typical four-year program.1
Then there are the courses in cognate areas such as discrete mathematics which, while foundational to computer science, don't touch on areas such as "secure systems". And there are all of the course topics required by accreditation, which aim for a broad overview of the field rather than concentration in particular areas – that's left to graduate education. Again, that makes sense; few of the students going into an undergraduate program in any field have a firm idea of what they want to do, and most of those will change their minds anyway. It makes sense for undergrad ed to try to provide a broad look.
There are certainly problems with CS education, at all levels, in the US. There are also a number of talented researchers and educators working on CS pedagogy and education in general, such as Mark Guzdial.
As with so many topics, people in the industry love to bemoan the state of CS education. They often don't seem to know all that much about it, though, much less be participating in efforts to improve it.
1Not all are four years, of course. My CS baccalaureate was a five-year program, but it was a mix of coursework and co-op study as an employee in the industry. It worked out to about four years' worth of coursework. And the four "years" are actually two four-month semesters each at most schools, so it's really under three years.
Well, we're talking university here, so students are prone to paying attention and not clown around, but it would probably be easier to find teachers if the levels before University had competent teachers and sufficient material and supplies to do the job properly.
A passion for teaching ? Kids growing up in the US are not given the environment to foster that kind of ideal.
"Well, we're talking university here, so students are prone to paying attention and not clown around"
Depends entirely on the local culture (and also, as in the UK, the kind of university - real or ex-technical college). In many of the latter at least, here in the UK there's a growing sense of entitlement to a 'good degree' as it's the certificate that counts (primarily as a passport to employment). Consequently many students do as little as they can get away with and staff tend to accommodate them. I have attended 'lectures' consisting of the lecturer handing out detailed notes and reading them to the class. Assignments containing literal extracts from said notes gained acceptable marks.
In the UK at least, there's now a culture of 'providing what students want' as if the university were a retail business.
Not sure about IT but I know a few uni science teachers in the US and the unis pretty much treat them like shit. I cant imagine the IT teachers in the US whose workload has doubled have received the pay increases that would attract others to the job, But I bet someones enjoying that more than doubling of departmental income.
Would you like young students/still growing immature adults to be taught the secrets allowing them to build a WMD?
The finer and broader secrets of AI and ML are many times greater a destructive force than that simple and very APT comparison. I Kid U Not. And that is probably why its finer and broader secrets are so expensive to buy and so much more expensive to deny and proclaim are a waste of defence resources and funding.
* .... Danger, Will Robinson!
About 15 years ago, maybe 25, I spoke with a bunch of Edinburgh University computer science students and asked what they were studying. AI, to a man. I burst out laughing because they didn't see a problem with this, even career wise. Nobody on databases? Nobody on processor optimisation? No language developers?
They think it is cooler than other specialities and will pay off big time any time soon. And those things are true for a very few of them, eventually.
A girl at a gig asked me what I did once, and was obviously contemptuous. "So you make the world go a little bit faster?"
"Not just that, more importantly I make a lot of low skilled people unemployed."
Next time a girl asked me, "I don't do anything, I never did nothing."
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
