Hacker News Reader

Not sure about your examples. I mean, when working as a programmer, we encounter tons of things we've never heard before but we do the research, we study what's needed to complete the task and we complete the task... e.g. it is quite easy to search about audio recognition - you encounter fingerprint, then you do the research...

Your examples make sense when you're isolated on some island without a connectivity, but that's not the case usually.

Also, I think companies like Shazam have separate people/teams in "science/research department" and "engineering department".

It‘s easy to google how Shazam works now, since they exist.

I‘m pretty sure that googling for „audio fingerprint“ or „algorithm for recognizing a song“ wouldn‘t have gotten you the answer in 2002.

You need to know the maths if you want to invent something new. If you just want to reimplement existing stuff, maybe you can get away without it.

(Inventing new stuff is actually pretty rare for programmers. I don‘t think I‘ve actually invented much, despite having worked as a software developer for more than 15 years...)

People tend to underestimate the age of working knowledge a lot.

Shazam wasn't an original invention, similar things were implemented in hardware in the 80s, and original research in this area done even much earlier.

Sure, though, as you go back in time, the probability that the environment you could get exposed to that knowledge in overlaps with academic circles probably increases.

Seriously, consider that the PARC people were working on perceptrons decades before they became truly feasible. There is almost always prior art to build on.

The most valuable engineering skill I find is the ability to say "I've run into a problem, I am not the first person to run into this, let me research how other people solved similar problems so I can understand my problem and it's edges better".

I'm pretty sure you'll get more than enough 90s, 80s or even earlier papers by searching "audio recognition", classification, etc...

Also, having studied math in the university and knowing where to lookup things when needed is one thing but demonstrating the knowledge on the blackboard during the interviews is another, especially if you're a mid-career professional and don't have a time nor willingness to allocate month(s) for preparation.

Yes, I know Google Scholar didn't exist in 2002, but there were papers anyway:

https://scholar.google.com/scholar?q=audio+fingerprint&hl=en...

Yet if you look at the engineering team at a company like Shazam, I'd bet a small minority of them actually work on song fingerprinting.

Shazam is an app company. Most of the people there will be working on UIs, serving requests fast enough, integrating with partner APIs, and other standard software development where they never once need to feed audio packets into libfftw or whatever.

Your other comment seemed to suggest that you feel the Boeing model should apply here, and the people who don't work on the math-heavy product core should be considered technicians. Is that the right interpretation?

They also need a decent database to find the song from a five second sample occurring anywhere in the whole up to say 8 minute song, for their whole song library

Also in some cases have access to not well known artists altogether.

I'm suggesting that a well-rounded team has a mix of skill levels and expertises.

In my experience this is where the ability to learn from literature would really help. In this specific case you’d encounter Fourier analysis pretty damn quickly if you started to read around the subject. Likewise whilst I think my formal teaching gave me a grounding I actually value the self-learning aspect from my degree more. The hard problems are the ones you’ve not heard the solution too before. The main thing a formal education provides is a leg up and hopefully a solid set of skills to help find answers to hard problems.

I can confirm that without specific education, reading around things like audio analysis and audio fingerprinting, Fourier comes up pretty quickly, and did before Shazam was a thing.

Honestly, I think tenacity and an ability to learn and move quickly are the most valuable engineering skills one can have.

>> Honestly, I think tenacity and an ability to learn and move quickly are the most valuable engineering skills one can have.

I agree. Nowadays the information is abundant and so easy to reach. One only needs to be able to understand and learn. From there, sky is the limit.

The only place where I think this doesn't apply is within an extremely specialized space. I think that very few engineers of that type are generally necessary, though. At the same time, I'd imagine that if you do have the academic background for extreme specialization, you're probably going to be able to demand a pretty high paycheck.

How many people on the Shazam team needed to come up with the fourier analysis? How big was the Shazam team? What did the rest of them do?

This is my problem with only hiring comp-sci to build software and get stuff done. They all want to hire the John Carmacks of the world, but Doom would have never happened without John Romero. It's a waste of a PhD to spend it porting low level driver software from one architecture to another, but that type of work needs to be done, even by companies like Google.

And honestly, there's a difference between the high-level big-idea fancy stuff that Computer Science focuses on, and the low-level, details of implementing things correctly and performantly. They are not the same skills, but that's all been lumped together under this software engineering profession.

'Without knowledge of FA I'd probably hack around for years with one inadequate ad-hoc solution after another.

Yeah, learning to read and appreciate papers and academic literature is an important skill.

Simply doing some reading before reinventing the wheel...

That's the skill.

On the flip side here, though, once you have that ability it generalizes very well. My original educational background might be in biology, but I know how to read a technical computer science paper and where to find them. It means that I may not have knowledge intuitively, but I know where to find and learn it.

I'm not at all against proper maths training because it can help with many aspects of an engineer's daily work.

How's your example different though from specific industry knowledge that typically doesn't come with your usual engineering curriculum?

An engineering degree is foundational.

For all the responders who suggest this kind of thing can be picked up on the job by competent professionals, in my years at Boeing, I never saw a single tech pick up stress analysis, calculus, kinematics, heat transfer, dynamic analysis, etc., on their own. Not one. Frankly, I felt that was a crying shame.

Nearly everyone needs a structured environment and program to learn that stuff.

> Frankly, I felt that was a crying shame.

It is. Given the right environment (which might include coaches or teachers) it should be possible to pick up knowledge about stress analysis, kinematics, heat transfer, dynamic analysis on your own.

Calculus is a slightly different matter because it's indeed more foundational than the other areas you mentioned. Still, I think calculus can be learned without a formal education programme, too. After all, most of it has pretty obvious areas of application.

> This is where having proper engineering/mathematical training pays off.

Rather, this is where personal research skills are needed.

I cannot imagine a programmer "hacking around" without researching existing solutions for obviously difficult problem.

To put it differently, none of the colledge Comp Sci "training" has yet been of any use in my 20 years of programming work.

> none of the colledge Comp Sci "training" has yet been of any use in my 20 years of programming work

I don't have a CS degree. But I have taken college CS courses, and have been able to apply them directly to solve problems I worked on later.

Probably depends on how they make curriculum. Some of the things I remember having been taught:

  runge-kutta
  analythic geometry
  turbo pascal (or one of his relatives)
  DFA/NFA

Some of the things I find useful at work:

  memory and cache behaviour on different archs
  Dijkstra's algorithm
  posix

I mean, chance of being lucky to learn exactly the things that will be useful in programming (as opposed to other less volatile arts) is a small one.

> job once of printing what was on the graphics screen on the printer.

I've had several variants of this, and the problem was not one of coordinate transform but either futzing with the API to actually get at the printer, or re-layout the items on the screen because transforming a 16:9 screen to a 1:1.6 piece of paper will look weird.

FA is the first thing that anyone who remotely googles for "identifying music technical implementation" pops up. The idea that someone couldn't figure this out on their own just because they didn't have rich parents who could afford to send them to some school is quite mistaken.

Put yourself in their shoes. You're you. You weren't sent to a school. What would you do to figure out the problem? Would you really hack around for years without success, or would you look up related material, read, study the problem, and check for existing solutions and scholarly articles on the topic? You know, all the things that a Real Engineer does.

And you'd get results quickly. Because you are a real engineer regardless of whether you went to a school.

I’d just like to add to all the people responding with “google didn’t exist back then” in the comments here, that libraries did exist. I found a great wealth of usable and applicable information in mathematical encyclopaedias, handbooks and software engineering books and journals (even 5 year old ones) in the public library where I live.

Let's say you did some research that suggested FA was the right solution. Now what? FA is not a freshman topic, it's a senior topic. You've got 3 years of math background you'd need to learn first to get to the point of being able to properly understand it (rather than blindly using it as a canned formula).

I'm speaking from experience. It's a generational gap, I think, which is why it's hard to understand. But it's true.

FA is using cosines up and down the spectrum to convert a set of samples (audio, image, etc) into a frequency spectrum (the magnitude associated with each cosine).

I didn't have to look that up. I just know it. And I'm no genius. This isn't nearly as difficult as it seems.

Especially because universities usually spoonfeed you information at an absurdly slow rate. There's value in that – I'd wager you could whip me if we have to bust out some paper and solve some integrals. But then again, I can just get a computer to do that hard work for me. I know what integrals mean and why they're useful, which is far more important than the rote memorization of techniques to solve them.

Now, if you want to argue that there are specializations that you can only learn in university, you'll get no disagreement from me. Distributed systems is a perfect example. On the other hand, I know those fundamentals too because I completed 6.824 on my own time. It's freely available.

And yeah, matrix math is a piece of cake too.

The world has simply changed. Information used to be valuable and guarded. Now all that matters is raw talent and motivation.

And credentials, sadly. But the world is slowly learning those aren't a great proxy for talent.

It's probably hard to empathize because you were well off enough to go to university. But you have to remember that intelligence is mostly genetic (probably), so talent can come from all walks of life. What would you have done if your parents had a massive divorce and someone squandered what little your family had? Are you sure you would have gotten a scholarship? When your parents never instilled the value of university into you from a young age?

> universities usually spoonfeed you information at an absurdly slow rate

I have to chuckle at that. I went to Caltech. A year of high school physics went by in about 4 lectures. I had to study about 40-50 hours a week just to keep from drowning. I would not have survived without help from the professors, TAs, and fellow students.

I don't disagree that a motivated, talented person can learn all that on their own. I've seen it happen, exactly once. The rest just employ ad-hoc hacky workarounds.

BTW, at the time Caltech had a specific policy that financial means would be no barrier to admission or attendance. They'd do what it took to make it work. I attended a very ordinary public high school in a working class neighborhood.

> matrix math is a piece of cake too.

But you have to know what it is to even recognize that it is the path to a solution.

Slow relative to learning the interesting bits. It took three years to get to FA? Seriously? That's a glacial pace relative to solving real world problems.

I agree you came out of it far better prepared, but drowning you in math for math's sake was merely hazing. Mostly.

Regarding your financial point: it's remarkable how hard it is for the upper echelon to empathize. It's not about the money. The money doesn't matter. It's the whole way of life leading up to college that kills you. The mindset. The motivations. The infighting. Being sent to some small religious school in the middle of nowhere because your dad thought religion was going to take over the world and it was super important for you to know a thing or two about god, then being tormented there because you're the only one in a school of 300 with an intellectual curiosity.

You think you would've stood a chance in an unfair fight? :)

Your mom and dad are owed a debt of gratitude. Family structure is the basis of power in life.

> glacial pace relative to solving real world problems / hazing

Sorry, I can't agree. My experience in post college real world problems was I was very well equipped to do so, compared with other newly minted engineers. And the techs who didn't have degrees weren't even on the radar.

> the only one

I was regularly assaulted in public school for being a nerd. Does that count?

> mom and dad

Mine were very special to me. But they were mine, and I'm heavily biased. Stepping back a bit, I don't believe they were exceptional, or better than the parents of my friends, etc.

> empathize

Get up and go comes from within, it is not installed by parents.

I wanted real bad to be a ham radio operator as a teen. I could not remotely afford to buy the equipment. I tried to build my own, but utterly lacked the skills, and could not get either the tools or the parts. Nobody helped me. I didn't know any hams, or anyone who knew anything about electronics. It was extremely frustrating.

For me, being at Caltech was very much like someone turning the lights on. I was around people with similar interests, classes were fun (despite being terrifyingly hard), lab equipment was all available, it was heaven. Nothing like public high school.

BTW, I was on partial financial aid from Caltech. They fronted me a loan, which I paid back after graduation. I'm sure they would have provided a full ride if my father couldn't pay anything.