Which computer science study field has the most potential?

Right now, it seems Artificial intelligence and machine learning is looked for everywhere, and it seems to have a bright future. However, I think bio-informatics is the one with the most potential, medicine and computer science are the fields with the most rapid evolution.

i'll list some CS fields

AI
Bio informatics
Computer graphics
Computer vision
Theoretical CS
Cryptology/cyber security
Quantum computation
Operations research
Data science


I'll be honest I'm soon entering a masters but I'm not sure what to specialize in so I created this thread to help me a little bit. Computer graphics looks fun but other fields seem to have a more promising future. Lets discuss CS fields.

Attached: 1495791415_New Picture (2).png (608x341, 92.84K)

AI/ML is popular now, but real time systems, embedded systems and computer architecture will be important as more computers enter the 'real' world. Whatever you do, you can't go wrong with a good foundation in mathematics (analysis, linear algebra, probability in particular).

real/complex analysis? Which topics in that could be useful for CS?

Geographic Information Science

hurr but google map

smallbrain.wojak

In Europe, Analysis encompasses calculus, multivariable calculus, fourier transforms etc. If you're making web pages, it might not be useful to you. If you want to work in computer graphics (I do not mean design), machine learning, anything with electrical engineering, or say signals and systems etc, then that will be helpful if not necessary. In the USA, perhaps Calc I,II,III will cover what you need.

Cont'd. Perhaps a good check to see if you're doing well is the introductory problems in Knuth's Art of Computer Programming Vol1. Are they too hard? Study more math. His Concrete Mathematics book is an easier alternative and expanded version of that chapter.

Strangely enough, the course which fails out most computer science majors is usually linear algebra. It's not at all a hard course, but it might be the first exposure to formal proofs students have seen.

Quantum Computing

Computer Networks
Database Management

Pick a field where the are not 400 million pajeets competing with you for the same positions.

I would say it is a combination of computation statistics, machine learning, (software) engineering and the sciences in which well-understood mechanistic models are available. Especially developing new statistical techniques built on machine learning to make the inference stage of more statistically powerful and computationally efficient. Think like Bayesian posterior inference without a tractable likelihood or marginal model. In my opinion, this will be an even more prominent topic due to its applicability to a complete sciences (medicine, biology, engineering, physics, ...). FYI, the literature calls this "likelihood-free inference".

So basically don't even try CS, right nocoder?

CyberSec is just going to get more important as IoT retardation increases. As long as there's valuable data stored on computer, cybersec has a future.
Definitely the field to get into if you want a secure job.

Congratulations on completing your 3rd online python tutorial.

I did not say he shouldn't take computer science, I specified he should pick a field which isn't oversaturated. You seem dumb, so I'll draw your attention to the OP question: "Which computer science study field. In the context of the question, field is referring to something in the realm of computer science.

Fortran.

There is no field that isn't over-saturated. There's >1.5 billion pajeets.

This is like being a shut-in and your excuse is to cite statistics on car accidents.

Go into botnet field, it's in high demand.

Seriously? Linear algebra was one of the easiest courses I took in college. I was even easier than a lot of the bullshit electives I took like "film comedy." Then again, I wasn't a CS major.

diversity and gender studies

I don't follow reddit memes. Is writing code some kind of achievement now?

yeah in usa I think calc 1 2 and 3 encompasses that. Definitely needed for computer graphics, the books about the subject use multivariable calculus.


pajeets write easy and retarded code, they don't specialize in a complex field.


cyber sec will always be relevant, but I don't think it will become more than what it is now.

interesting point, I'll look that up

How long until this isn't a meme?

AI is the root of all computer science.
All good operating system developers were into artificial intelligence.

I would say, like, in 15 years maybe.

OP is taking undergrad courses, so these are mere concentrations. No one thinks of an undergrad as any sort of specialist.

Yes I agree but I, OP, am planning to take graduate courses, so I use the term specialize. I think someone dedicating a 2 years masters to study to a narrow subject in CS would be specializing in said subject

We basically hire undergrads at the faculty to do pajeetwork. It is a nice opportunity for them though, there are proper projects, which allows them to "specialize" in the topic. Contrary to what you would get in the average company.

What do you think is interesting, what are you good at? More importantly, where do you see yourself in 5 to 10 years?

almost every topic in CS, honestly

if I look at my grades, TCS. And I'm learning some computer graphics basics in my free time right now

computer graphics is looking fun right now

I think AI has more opportunities though and my city is an internationnaly known hub for machine learning research

What do you find enjoyable, like, you could lose a complete day upon without realizing the passage of time.

I personally never look at grades, it doesn't say anything about a person's creativity or research-potential. Typically because most people which are like this, - really - hate to write formal reports which is what most academic personnel cares about.

If you only do it for the money, believe me, you will be depressed. Because the industry (especially Deloitte etc) are filled with retards which apply the techniques incorrectly. Also, consider that the acceptance rate into Google Brain, DeepMind, MAI and TAI will be quite challenging, unless you are a female or PoC (not kidding, I even know of workshops at NeurIPS which reject high-quality papers to diversify the pool, but that is a different story).


Outside of the major you are considering, is there something you would like to achieve? Some kind of greater goal, or wanting to work in a certain environment with a certain goal in mind?

if I forget about money or grades then computer graphics sounds like what I would study more.


There's so many unsolved problems in computer graphics, if I could find a way to solve one of them that would be great, but to reasearch new ways to simulate things I think a ph.d is needed. Otherwise, I don't have huge ambitions, if I can manage to work on an interesting project in a good company, I would be happy. However in order to do that, I would need to be in a good position to do meaningful work. And like you said, some interesting companies/labs have a very low acceptance rate so I really need to git gud. (I'm a straight, white privileged male)


Thanks for the tips, user. Here's a nice rat for you.

Attached: cute-pet-rats-26__880.jpg (880x677, 43.21K)

Just one last recommendation. git gud at statistics and make yourself comfortable with the notation. For instance, Combine it with your simulation setting, since you are interested in graphics. Say, you have a stochastic complex simulation y ~ f(x) and now you want a distribution over x given y p(x | y), how would you do that. Train yourself on projects like this!

All companies and institutes are basically formalizing actual problems in some probabilistic framework. That's why you see frameworks like Pyro etc.

Good luck.

Speaking of statistics let me disprove my 1.5 billion pajeets statement earlier.
Let's assume the average IQ of programmers needs to be 115 with SD 15, I'm guessing at this because I couldn't find a site listing SD with their claims about programmer IQ.
India has about 20 million of these, putting it at 4th in the world.
So no there isn't really a giant flood of high quality pajeets. However, China is just ridiculous. At >=130 China has 73mil people Japan, in second place, has 6.5mil.

Probably the best field will be of moderate difficulty. Anything super high level will be chink central and anything low level will be pajeet central.

I have a gutfeeling it is less. Where exactly did you find the information regarding the IQ distribution of India (assuming Gaussian?).

You can't just assume equal IQ distribution across different countries with vastly different living standards and levels of education.

You think there's more smart Indians than average Indians?
fucking lmao India the place where everyone is above average

That's why I said "assuming Gaussian" faggit.

if chinks are so smart, why are the white boys still dominating tech?

I think the Chinese artificially inflate their domestic IQ. In the West, we see the cream of their crop in immigrants which is why they perform well compared to the average.

Jew media also likes to push the idea of Chinese brilliance, "Asians are good at math!" As yet another way of denigrating white accomplishments. Always tell jews how great Chinese are at physics in return. OYYYYY BUT MUH EINSTEIN, MUH FEINMAAAAN! I think the nation that really does punch above their weight when it comes to Physics is actually Denmark.

No that's why you said "I have a gutfeeling"


hmmm

We're not only importing high IQ people. That was more likely to be the case up until the late 1960s, after which immigration was opened to the third world.

What you're referring to is blamed on a host of factors, miscegenation, integration (how do you learn physics in the company of niggers?), shoehorning diversity into leadership positions, rewarding the lowest common denominator, etc.

With integration, I realize that doesn't change the IQ, but it creates a loss of realized potential. If smart people are hindered, they may miss out on opportunities which could lead them to find a high quality mate.

You should specialize in the field you like the most.

This. Doing something you aren't interested in is not worth the amount of money you can theoretically make. Nobody's gonna give you all that time back and you can't buy it back either.

I'm doing what I like

Ff you can get away with it and it doesn't cause you distress in some way, congrats. 20 years down the road from the day you die nobody is gonna give a shit if you were a useles neet, wrote some really good programs for your corporation or even were the fucking CEO of Apple. Spend your time in ways you like, because it's limited no matter who you are.

Hindi

emedded systems isnt really CS. computer engineering is emedded systems and computer architecture. to understand embedded systems you need to now electronics really well. so that might put you off.

I personally consider the theory of computer architecture to be CS. I consider it an application of boolean logic to implement a computing machine (Turing complete machine).

sure you may go over some of it in CS classes,but you dont go as depth as they do in computer engineering classes. were you might have mutliple classes about it. they probably should go over it more in CS classes the same for assembly but what can you do.

Pick something difficult and boring, that means you'll make bank and you'll have less competition for work, and less idiots in your field. Something like writing firmware for hard drives

that sounds fun though

That's the sentiment being expressed here , but the Python programmers got upset.


Pretty much everything once you get into higher level topics, mathematical literacy is an absolute must. You can confirm that by looking at image processing, machine learning or networking papers/textbooks aimed at the advanced undergrad to graduate level.

How about just learn to code and >>>/hydrus/ ?

no

nocoder spotted

You're ass backwards.

Computer Science holds the science of computing as it's subject matter. Theoretical computation machines (e.g. Turing machines, Von Neumann's work back in the day), algorithmic complexity, the implications of concepts, information theory, &c. I don't think it's a proper science but, akin to the physicists (and mathematicians if you take Goedel's view), the computer scientist's goal is to push the envelope of what is known about the 'theory' of machine computation. They're discrete mathematicians with an emphasis on the math which may be executed by our machines. Plenty of overlap with other sciences; a well known case being quantum computing. Knowledge of QM was necessary in order to build transistors in the first place, but as time went on and mathetical physicists and mathematicians began to play around with the possibilities of computing with the superposition of bits some practical applications began to arise (e.g. Shor's algorithm). They don't care much about the physical implementation. Questions of silicon vs germanium and the geometry of a chip's litho mask don't matter to them.

Computer Engineers engineer things. Engineers are one step down the epistemological line from scientists. They're concerned with applying the body of knowledge of their field rather than developing it (though our labels are imprecise and often scientists/engineers end up temporarily switching roles). Their domain spans both hardware and software, and they seek to realize EE/physics/material science/CS concepts in the real world.

Re: computer scientists: I actually do tend to think of them as engineers rather than scientists. Their work is a bit closer to practicing science than the computer engineers, though, and it isn't a clean distinction. Engineers often use scientific methods to solve implementation problems and scientists must often engineer devices which allow them to test their hypotheses. In any medium or large project they will almost certainly be working together.

/end pedantry

AI/ML are applicable to pretty much all of the fields you listed. Computer vision is the application of machine learning to a specific set of problems.

good job writing that

Doing data science and analysis for a living now, part of it (first quarter of this year) was building a data warehouse. Initially started as a software dev (mainly. Net - C# and F#).
Just pick one and go with the flow. You don't really know what you like until you work with production code.

Do you like what you do?

tranny computer science (tcs)

I do, yes. Never thought I'd work on 10 TB RAM clusters.
That being said, if I ever outgrow it, I am fairly comfortable in my skills to start doing something else.

Infosec. All those antivirus companies need to keep writing better viruses or else their products just won't sell.

...are you implying antivirus companies create viruses that their software is immune to make other antivirus companies fail so they win the competition?

He's saying that they write malware so that their market has value. No viruses = no need for caution/protection.

That sounds as idiotic as policemen creating crimes specifically to increase their social value.

How many police departments have justified additional spending by calling in 'reports of a suspicious package.' Oh no, we need a really fancy robot now!

Zero.

s/social/financial

No pajeets can actually code though, it's a meme. While companies persist in being retarded & hiring them, they inevitably hire more non-pajeets (ie, White guys) to come in and fix the mess, cost far more than necessary in time and money. It's a shitshow, but it's also why any coder with more than a few months of experience & the slightest ability to interview is never unemployed.

anons, please tell me it's east to get a job as a cryptoscientist working for a comfy government

Teledildonics