Ask HN: If you could go back to study any CS-related field, what would it be?

[offbytwo]

I'm a sophomore in college, and feeling pretty bogged down by the not-so-relevant required courses at my school. I love the CS courses but I keep finding myself looking at entry level code monkey jobs and thinking of dropping out. I work part time as a developer right now and I enjoy working far more than doing any of my homework, so this is something that is on my mind a lot.

What are some lesser known areas of CS that would be worth studying while I have the chance? I would say the subjects that excite me the most are Machine Learning, p2p tech like IPFS, UX-design, and alternative computer-interface things (like brainwave sensors, VR, and that jawbone thing from MIT that was posted a few weeks back [1]).

[1] http://news.mit.edu/2018/computer-system-transcribes-words-users-speak-silently-0404

[jimmies]

As a former-dropout myself [1]: You might some of the courses utterly stupid now, and you might have many doubts about the usefulness of a degree, but I think ultimately a degree is very very much worth it, both intellectually and for logistics reasons. A degree will open doors to you, for example, most jobs will throw your resume right away if you don't have a degree. Some countries won't allow you to immigrate if you don't have a degree. There are many dreams that require a degree.

The only reason that you can justify dropping out is that either (1) you think you can't possibly learn anything useful from the professors that are teaching you and you'll rebuild/repay what you didn't learn one day (and you better have to have a good answer when you'll do that right now), or (2) when you have a grand startup like Bill Gates or Mark Zuckerberg. But I don't think that's why you're wanting to drop out now. So don't drop out. Keep pushing.

For me, I wish I learned assembly, kernel development, stats and machine learning. First two because I love to, the latter two because they are useful.

I am now almost finished with grad school, and I feel like I know nil. But in a very Lao-tzu way, I think the biggest enemy of mine is myself (the willing to sit my ass down and learn), not that these can't be learned by myself. Lately, I think I somehow I overcame that problem and was able to read, learn and make a lot of stuff on my own. I think the same thing can be said about anyone who had the patience to get a degree as well: It means they are willing to deal with things they don't totally enjoy to get what they want. As Lao-tzu said, patience is a good virtue by itself...

1: If you need to verify, read the entry called crankshaft #2 on my blog on my profile.

[washadjeffmad]

I remember signing up for a course in mass transfer that we joked would have been better titled bubble science. It seemed so obvious at first, but as we progressed, the math and concepts became more vast and inscrutable to me, and by the end I'd learned so much about something outwardly so simple that I felt that I'd barely scratched the surface of the topic. But that's not really an unusual outcome for learning, right?

Pursuing a technical education is tricky because those interested often have an elevated baseline knowledge and want to jump ahead without relearning fundamentals, but it's often those fundamentals that cause growth to suffer later on. Realizing that you're actually struggling with algebra while you're taking mv calc is a big eye opener, and realizing that you get the basics and applications of certain implementations of certain technologies in the first two weeks of a course can feel redundant and insulting, but that's because it's hard to gauge or trust that there's more to things beyond the limits of our understanding, not because the material is unworthy.

Overcoming that is humbling, and that can put people eager to get a start on making money because they're already slightly better at something than the population at large at odds with the goals of higher learning, but it's a necessary part of our growth and perspective.

People who grew up being told how smart or special they are can have a harder time with this, and I know it was pretty embarrassing for me when I realized early in my adulthood I was much closer to the "kid who's good with computers" category than an actual "IT professional", despite being able to successfully complete contracts and make money from what I was doing. Those experiences helped me re-evaluate my approach and get out of the "I'm already awesome, why would I need to do more" mindset. Had I not realized that, I might have stubbornly stalled out thinking I didn't need anyone else while the world passed me by.

[CalRobert]

"you think you can't possibly learn anything useful from the professors that are teaching you"

Of course, it is possible that you will learn from the professor, but that you would learn _more_ by doing other things with four years of your life. University ahs no monopoly on knowlege.

Also it's disturbing what starting life with a gigantic, crushing pile of debt can do. (It doesn't have to be this way, but I've seen plenty of even quite intelligent individuals go this route and regret it).

If you can, take advantage of things like junior colleges. They're amazing. I owe my career not to my degree (pointless, though people at least tend to say "well you can't be an idiot" when they see a physics degree), but rather to what I learned taking a few night classes at Santa Monica College (including assembly, which was good fun) while working a day job and doing some projects for fun.

[mehrdadn]

> For me, I wish I learned assembly, kernel development, stats and machine learning.

Did you take compilers, learning e.g. parsing theory? If so, are you happy you did or did you feel it's skippable? If not, do you wish you had taken it?

[sp527]

For the sake of registering a counterpoint: I disagree entirely that things like kernel development or assembly (and let’s throw in architecture, computability theory, and all that jazz for good measure) are even remotely useful in software engineering. You’ll forget most of it and personally I don’t think it will even meaningfully alter your performance over the long term.

Knowledge that is acquired but not routinely recalled or applied will atrophy.

Sometimes you can make the argument that it’s worth your time to satisfy your own intellectual curiosity and I can understand that. Where people misstep is in thinking all knowledge is created equal.

I used to rationalize forays into theoretical material as holistically improving my capability as a thinker. In hindsight, it’s obvious that was bullshit. There are much more efficient ways of turning yourself into a good thinker that are more directly relevant to how things work in the real world.

The other thing I realized (and this is more specific to me), is that if I were to give myself the luxury of diving into knowledge for its own sake, I would choose a topic in the natural sciences, like physics or astronomy. Computers are interesting, but the theory surrounding them doesn't do much to help explain the nature of our reality, which I personally find much more fascinating.

If I could go back and redo my education, I would try my best to focus on a combination of:

(1) The most pragmatic courses in CS. IMO, the most useful ones beyond the intro courses were data structures and algos, distributed systems (project-driven), OS design (writing a simple OS), basic prob/stat, and intro ML (you do not and never will need deep anything, unless you decide to specialize). You could cover all of that in about a semester and a half tops.

(2) Projects out the wazoo. Real ones. Ideally motivated by a real problem and birthed into the world with all the messiness that entails, and iterated upon until they create real value for someone. You'll learn a stupid amount along the way.

(3) Through some combination of courses, reading, and projects: scripting/automation, API design (easy), modern web dev (project plus lots of Googling and learning to accelerate learning by relying on others), mobile app design (same approach as web dev), PaaS via AWS or GCP (or bespoke), basic security, AMQs, orchestration (at least Docker; maybe Kuberbetes), proxying (uses of Nginx) and UNIX/Linux networking fundamentals, metrics and analytics (with an emphasis on learning the value of instrumenting a system/product/business and using the feedback to improve it), databases (Postgres at least; become super proficient at SQL), basic UI/UX design principles, software engineering best practices (from simple things like KISS, coupling, testing, all the way up to reliability, availability, maintainability, scalability, and good decision-making, particularly with respect to knowing how to achieve a sensible balance between time, cost, and quality).

I’m missing a lot, but in short you should know every technology function required in a modern company at least at a basic level. Some people call this "full stack".

If you want a lasting career in tech and you don’t plan to specialize, then this is the way to go. The merits of being a specialist vs a generalist are debated all over the place. Thiel will tell you to relentlessly focus on one thing and ‘vertically integrate’. Scott Adams will tell you to get very good at two or more things and then combine them, since becoming the best at any one thing is extremely hard.

If it’s not obvious, I chose to be a generalist. If I had to explain why, it would be because: (a) I don’t like the risk of committing to one thing (“blockchain engineer" seems like a dubious track, for example), (b) I get bored easily, (c) specialization often but not always seems to lead to myopia, which is cancer in any enterprise; this is hard to explain but you’ll know it if you ever see it: everyone operating in their own silos, incapable of cross-displinary thinking, lacking empathy for the nature of what other people do, pervasive groupthink, arrogance (d) if you’re not good enough to be a top-tier specialist (I'm not), then the way you maximize the value you can create and that you can get paid for is to be an exceedingly useful generalist, who can think across organizational concerns and boundaries effectively.

(4) What Charlie Munger calls “remedial worldly wisdom”.

The most appalling failure of our education system is that it produces people who can take a test but can’t think independently, let alone innovate.

Some of us software engineers get to thinking we’re hot shit. We're not. For one simple reason: what we do is almost always deterministic. Someone has done it before and written it down so that you can do it too. At worst, you have to tweak something a bit to make it work for your situation.

In the real world, nondeterminism drives novel value. In other words, everything wrapped around the lines of code you write is what's important. That means you're going to be hard-pressed to make a dent in anything if all you can do is write code.

Thinking well is a broad subject and you’re going to have to tackle it on multiple fronts, probably for the rest of your life. The most important thing by far is behavioral psychology. Do whatever you possibly can to grasp it. Additionally: systems thinking, philosophy, basic accounting, very basic economics (as soon as they say “Solow Model” run away; ideally well before that), some history. Poor Charlie's Almanac is a good starting point for much of this. It'll help you appreciate why this is important.

You should also know how to apply math to solve any problem you run into that falls short of involving calculus or advanced prob/stat. In a perfect world, you would know how to apply calculus as well, but the opportunities to do so are so few and far between that you likely won’t remember it beyond basic differentiation/integration in the long run (or at least I didn't since I have a poor memory, but that may not apply to you).

(5) Read The Lean Startup. And expand out. Be careful since there’s a lot of garbage in the business genre. Others I can recommend: The Phoenix Project, Lean Analytics, the first part of The Startup Owner’s Manual (the latter two parts only if you ever get past the first stage of building a company). Even if you never choose to work on a startup, it’s the same kind of thinking that will enable you to generate outsized value in any organization. Good decision-making offers at least an order of magnitude better value per unit time than writing code. You will get in the door by writing code. You will get up the ladder by making good decisions.

When you read books, get paper copies and write in them: underline, take notes in the margins, drop in some Post-Its to mark really good sections, etc. If you read a book that really resonates with you, then go further and write up notes on it afterward. Even just underlining a book is ridiculously useful. Underlining alone can allow future you to skim through what you understood to be the most important parts of a 300 page book in roughly 10 minutes.

All of the above may seem like a lot. And honestly, it would all fit in easily if I could swap out the less useful required parts of my CS degree. But that won't be viable until universities offer that option and companies stop thinking a complete CS degree is the qualification they should be targeting. Until that happens, the onus is on you to not let your "schooling interfere with [your] education."

[kaeluka]

That blog entry is well worth reading. It put a smile on my face! :-)

[LocalMan]

I dropped out in 1971 for reasons similar to yours. I am now retired. I've supported myself by writing software since the 1970's.

Dropping out has cost me several hundreds of thousands of dollars. I don't know how many. Lots of companies pay more just for having a degree.

Consider transferring to a cheaper and/or easier school. You'll have more time to yourself, which is often a good thing. Unless you're actually depressed.

If you're interested in CS, math is almost always useful. I wish I'd had more. Most of the topics you list are research topics only available to research-level academics.

Changing schools is the one thing I regret not trying. Another thing to consider is to change girlfriends. Or find a better one. That can certainly impact your overall view of life. In other words, don't ignore the social aspects of choosing a school.

In the 70's through the 90's I had to contend with management that was often quite stupid. And quite often did not even know what a computer really was. You can avoid situations like that more easily if you have a degree.

[gitgud]

No-one has mentioned Cryptography. Although it seems boring, the applications are almost ubiquitous. It's used everywhere in internet connected systems, secure software and data storage (to name a few).

It would seem that would get to work on the large and important things if you pick up cryptography as a skill.

[greenhouse_gas]

And also, its one of those things that's hard to learn on your own. On the other hand, do they cover defense against practical attacks in uni? As in, will you be told to roll your own crypto, and have the professor show you the timing attacks, etc.? Because honestly, the theory you can pick up on your own, in industry you just use a library, and I don't know how nice will OpenSSL devs be if you just show up as a crypto noob on their doorstep asking for mentoring.

[ericmcer]

require('bcrypt').hash(pw);

My main hesitancy with the really domain specific courses offered in college is that the careers associated with them are all or nothing. Adding stuff like machine learning/AI, cryptography, etc. to your skill set as a standard software engineer is super hard to do in a practical manner.

[gnodar]

I would take some Electrical Engineering classes, maybe even minor in it. Given an abundance of time, I would also take some higher level math classes. Based on your interests a broader exposure to these fields, which are strongly related to CS but only lightly-to-moderately covered in a standard CS curriculum, can only help.

For me personally, I would do a deep dive on distributed architecture, which you may also be interested in given your interest in p2p.

[mehrdadn]

When you say EE, do you mean signals/optimization/control/etc. or do you mean circuits/device physics/RF/etc.?

[edejong]

The ones that will help you long-term: management, requirements analysis, distributed aystems architectures.

Technologies come and go and you’ll be autodidact during your career. Management, however, will help you identify how to become more effective, regardless your actual posistion.

[quintes]

This

[nojvek]

Compilers - every programming language uses this.

Databases - Btrees, indexes, distributed key value stores. The world runs on databases.

Graphics - if you like games, how a 3D scene is rendered, photo realism with Ray tracers, GPU pipelines and OpenGL

AI and Machine Learning - tons of fascinating problems and algorithms.

I’d say focus on the basics. Hash tables have remained mostly the same since they were invented. C still uses pointers, the basics are fundamentally so powerful that they are kind of eternal.

When I hire someone, i’m looking for someone with strong fundamentals. They understand the basic datastructures, algorithms and how a computer works.

[your-nanny]

I regret not taking linear algebra (not required by most programs, but very relevant to what I do), computer graphics, and software engineering methodology. If I had taken these classes, I would not have taken other interesting and useful coursework that I would regret not having taken. Hard to measure which path would have minimized my regret. Possibly none, because it may be that degree of regret isn't a function of the path you take.

[thisone]

Same as with linear algebra. It was never part of my course work. Differential Equations, yes. Linear algebra, no.

I plan to teach myself one day. Maybe I'll start today

[Endy]

Being honest, if I could study any CS-related field, it'd be pretty simple - advertisement technology management. I'd want to know how to get in the game of stealing people's private data for money; it seems that's the wave that's going to be cresting soon - and people always want to know more about other people.

[manish_gill]

One thing I've been always fascinated with was the Symbolic Systems course at Stanford. It was a fascinating curriculum involving a mix of Computer Science/AI, Philosophy and Linguistics. I've tried searching for material online, but it's a course taken by only a few people every year and there isn't a whole lot of information available online. If someone reading this studied that/or knows about the methodology/syllabus, I'd love to delve into it!

[Animats]

That's "traditional AI". I once took John McCarthy's course, "Epistemological problems in artificial intelligence", also known as "Dr. John's Mystery Hour". It was about how to hammer the real world into predicate calculus. That turned out to be mostly a dead end.

[jahewson]

I felt exactly the same way and made it through my undergraduate degree, worked for 6 months, realised that being a coding monkey was going to consist of repeating those same six months for decades and promptly returned to school, ending up with a PhD.

The best advice I can give is to find a mentor - someone who captures your imagination. Most professors are desperate for enthusiastic students to do stuff for them!

And yes, do more math. You need people around you for that.

[fyfy18]

Out of interest, what are you doing now you have a PhD. How does it compare to being a code monkey in terms of a career (salary, hours, prospects)?

[msoad]

Databases. Those are the most fascinating and complex programs I know but I have no idea how exactly they work.

Most of complex algorithms and data structures are used in databases.

[Animats]

Databases are conceptually simple. It's making them go fast reliably that's hard. Naive databases are "Lock everything, take search criterion, run over all records, return result, unlock." (DBASE II in the DOS era actually worked that way, and that's what people using Excel as a database are doing.) Now make it go fast and make it reliable despite failures while updating.

[baus]

I agree with everyone who says focus on math and stats. The best programmers I have worked with have come from a math background. This is an area I greatly underestimated in my education, and if I went back to school now it is where I would focus

[muzani]

I made the perfect choice when starting with Android. Great job prospects, great pay, allows me to go in and out of a lot of fields.

But if I could try over, I'd focus more on small skills instead of breadth. Full stack is nice to know but ultimately not useful. Anyone can learn to program something over 3 months. That doesn't necessarily make you valuable.

What makes people valuable is being better than other people at a skillset. Like right now we really need a good AngularJS (1) programmer, but that's hard to find.

"UI/UX guys" are a dime a dozen, but what's extremely valuable are the ones who can prototype quickly, write their own CSS/HTML. These guys will be core to any group.

There will always be new, sexy tech. The hard part would be coding the algorithms. The guys who are cashing in on e-commerce know their Big O. The guys who are well paid writing code for Uber know their algorithms. The rest will change, and will either be reading documentation or copy paste.

[kyleperik]

I feel as though many of the comments here are throwing out fields because they're interesting. I think college is a very long term investment that will change what you'll be spending your valuable time on for the next 4+ years.

I agree with many comments that suggest pursuing something more "meta" like management or architecture that will help no matter where you land in 5-10 years. Although personally, I'd say experience is practically the best education you can get.

[jimpudar]

Not exactly a "lesser known" area, but computer architecture and operating systems. No matter what you are doing with computers, no matter what high level language you are using, you will eventually need to understand what is happening "on the bare metal".

[ikeyany]

Computer architecture was my undergraduate focus and sometimes I'm not too sure what's happening on the bare metal. You'll want to take VLSI, semiconductor physics, and a course on compilers to get the entire picture.

[floatingatoll]

Factory operations. In absolute seriousness! If you want to get years ahead in Ops, learn queuing theory and applications years ahead of your CS peers :)

[jimpudar]

I'll second this - queuing theory is extremely relevant in all sorts of places. If you haven't studied it, you might not know when the concepts will apply.

It's applicable to all sorts of things like event loops, job queues, network packet analysis, even database access.

[brightsize]

Geospatial technologies. Cartography, visualization, remote sensing, geospatial databases. If you're good with Python you'll go to the head of the class. My impression is that most students and practitioners in GIS don't have CS backgrounds and thus programming and relational/geospatial database (e.g. PostGIS) skills are in demand.

[tytytytytytytyt]

I would take all of the electives I could, if I could go back in time, even if it required staying an extra year. Graphics, networking, databases, ML, everything. I would even go for the masters, really. The undergrad degree is in a sense the prep work for the really interesting stuff.

[bennyp101]

Something that I looked at doing recently was getting into law around the internet/AR/cryptography etc.

It seems that there are a lot of people currently making laws and rulings on things that they don't understand - and there is going to be a lot of change coming soon with the way technology is going.

Unfortunately I can't afford the time or money to do it now (I think it was like 6 years to just get qualified) which is a shame.

(I never went to university or even finished my A-levels, I just went straight to get a job at 17, looking back 20 years later, it hasn't hindered me in anyway, but I do think that going would have had a positive effect and maybe changed my career)

[xfz]

Don't drop out; you'll spend your whole career trying to push open closed doors (it gets easier with many years' experience or during an acute skills drought, but the issue never goes away completely).

Study whatever you enjoy most. It'll be easier for you to excel that way, and you'll still get the all-important degree.

Once you get your career underway, continue to learn and work on whatever interests you most; be prepared to continually learn and adapt over the decades. New technologies and ideas will come along that haven't been imagined yet, while some of the stuff you study at uni will be surprisingly relevant later.

Best of luck!

[LocalMan]

As you get older, it gets harder to learn new stuff.