/sci/ - Science & Math » Thread #11285156

90KiB, 760x427, 21454.jpg

View Same Google iqdb SauceNAO

Computer science is actually the most intellectual field

Anonymous Tue 07 Jan 20:19:36 2020 No.11285156 View Reply Original Report

Quoted By: >>11285201 >>11285338 >>11286840 >>11286850 >>11286862 >>11286877 >>11286907 >>11286927 >>11287064 >>11287081 >>11287875 >>11287880 >>11288899

>Algorithms, data structures, and cryptography require some knowledge of pure and discrete math
>Human-computer Interaction requires some knowledge of psychology
>Computer architecture requires some knowledge of engineering
>Artificial intelligence requires some knowledge of mathematics, philosophy, and linguistics
>Machine learning requires some knowledge of statistics
>Computer graphics requires some knowledge of topology and geometry
>Modeling and simulation requires some knowledge of physics, chemistry, and biology
Only polymaths can study CS

Anonymous

Anonymous Tue 07 Jan 2020 20:24:32 No.11285165 Report

Quoted By:

Who cares which is the most intellectual. Im sure sociology can be quite intellectual, but its not marketable at all. I went to college and got a degree that was marketable, I read about math and physics in my spare time. Contrary to popular belief, you can study stuff outside of a classroom. But a degree is valuable as a signal to employers, and self education is not in any way equal to a degree when it comes to signalling.

Anonymous

Anonymous Tue 07 Jan 2020 20:40:59 No.11285201 Report

Quoted By:

>>11285156
based and matrix-pilled.

Anonymous

Anonymous Tue 07 Jan 2020 21:50:45 No.11285338 Report

Quoted By:

>>11285156
CS majors don't even shower.

Anonymous

Anonymous Wed 08 Jan 2020 14:52:36 No.11286838 Report

Quoted By:

ok zoomer

Anonymous

Anonymous Wed 08 Jan 2020 14:53:47 No.11286840 Report

Quoted By:

>>11285156
based

Anonymous

View Same Google iqdb SauceNAO QVfeXSb.png, 45KiB, 593x210

Anonymous Wed 08 Jan 2020 15:07:53 No.11286850 Report

Quoted By:

>>11285156

Anonymous

Anonymous Wed 08 Jan 2020 15:11:45 No.11286853 Report

Quoted By: >>11286868 >>11287953

CS is a pajeet/aspie degree that gets outsourced as soon as companies get the chance. The only decent outcome of a CS degree is software engineering, and only about a third of CS majors do that, maybe less. The rest become code monkeys which could have been done with coding bootcamp. CS is ECE for those that are bad at math.

Anonymous

Anonymous Wed 08 Jan 2020 15:20:36 No.11286862 Report

Quoted By: >>11286909

>>11285156
So how are art majors working at Google as SDEs after completing a bootcamp and a few LeetCode questions?

Anonymous

Anonymous Wed 08 Jan 2020 15:25:05 No.11286868 Report

Quoted By: >>11286965

>>11286853
Does it have to be a degree as opposed to a field of study tied to computation? The most mathlets can accomplish is get a degree. The most programmers can accomplish is causing the Singularity.

Today's math sits on broken foundations, a shadow of its true self. Computers are the cure for that.

Anonymous

Anonymous Wed 08 Jan 2020 15:34:46 No.11286877 Report

Quoted By:

>>11285156
Agreed

Anonymous

Anonymous Wed 08 Jan 2020 15:57:23 No.11286907 Report

Quoted By: >>11286941

>>11285156
Can any of you experienced anons tell me which Data structures and algorithms course is better?

https://www.coursera.org/specializations/data-structures-algorithms

OR

https://www.coursera.org/learn/algorithms-part1

OR

https://www.coursera.org/specializations/algorithms

Anonymous

Anonymous Wed 08 Jan 2020 15:58:32 No.11286909 Report

Quoted By: >>11287886

>>11286862
They are essentially tradesmen and they don't understand the underlying theory of computer science. Just like how chefs don't understand chemistry and mechanics don't understand mechanical engineering or physics

Anonymous

Anonymous Wed 08 Jan 2020 16:05:11 No.11286927 Report

Quoted By:

>>11285156
100% true, and it's one of the few degrees that will actually force you to think and there's no room for dogma or subjectivity

science and engineering still have some dogma involved

Anonymous

Anonymous Wed 08 Jan 2020 16:11:44 No.11286941 Report

Quoted By: >>11286961

>>11286907
Pls help

Anonymous

Anonymous Wed 08 Jan 2020 16:25:12 No.11286961 Report

Quoted By: >>11286968

>>11286941
I can't comment on the specializations as I have no experience with them, but the lectures by Sedgewick were quite good and his explanations were clear when I went through them a few years ago. I'd recommend that one.

Anonymous

Anonymous Wed 08 Jan 2020 16:26:27 No.11286965 Report

Quoted By: >>11286982 >>11287056

>>11286868
If computers sit on a foundation of math, and math sits a broken foundation, how can computers cure that?

Anonymous

Anonymous Wed 08 Jan 2020 16:27:48 No.11286968 Report

Quoted By:

>>11286961
Thanks

Anonymous

Anonymous Wed 08 Jan 2020 16:42:15 No.11286982 Report

Quoted By:

>>11286965
Do not worry about Norman, he is trolling as usual.

Anonymous

Anonymous Wed 08 Jan 2020 17:48:50 No.11287056 Report

Quoted By: >>11287257

>>11286965
> how can computers cure that?
By making computation the essence of proof as opposed to meaningless symbol pushing over uncomputable quantities.

> If computers sit on a foundation of math
It is the opposite way around, computation should be the foundation of math. It has to be for math to be rigorous.

The way mathematics is done now makes it the enemy of reason and understanding.

Anonymous

Anonymous Wed 08 Jan 2020 17:58:07 No.11287064 Report

Quoted By:

>>11285156
Yet they still can't figure out how to lose their virginity

Anonymous

Anonymous Wed 08 Jan 2020 18:14:00 No.11287081 Report

Quoted By:

>>11285156
Rate my Tree.
[code:lit]
#define tree
[/code:lit]

Anonymous

Anonymous Wed 08 Jan 2020 20:05:53 No.11287257 Report

Quoted By: >>11287279 >>11287445 >>11287448 >>11287453 >>11287457 >>11287513

>>11287056
Differential geometry mathfag here. I have some questions
How do you prove that an algorithm is correct?

Can you explain this more
>The way mathematics is done now makes it the enemy of reason and understanding

I took some applied math courses and some other physics courses too, but in all of them I was told that algorithms are pretty much useless for understanding, just forcing a computer to print an output is not science, you have to look for interpretations for the results using the theory, so in other words CS will never help me to give an interpretation to my results

Anonymous

Anonymous Wed 08 Jan 2020 20:18:52 No.11287279 Report

Quoted By:

>>11287257
it will help you understand the overall process, CS is like a gate to different specific areas, good entry point but you can't stop at it

Anonymous

Anonymous Wed 08 Jan 2020 21:44:59 No.11287445 Report

Quoted By:

>>11287257
> How do you prove that an algorithm is correct?
Formally, you can do it using type theoretic PLs: https://softwarefoundations.cis.upenn.edu/

You implement a program in a language with sufficiently expressive type system, and prove properties about it. Those proofs are also programs. So you are essentially writing programs about programs if that makes sense.

This is actually quite difficult and nobody really does it for regular programs. There are some lightweight alternatives to formal methods such as randomized testing where you write a generator and then test a program on a wide variety of inputs: https://fscheck.github.io/FsCheck/QuickStart.html (see also vol 4 of Software Foundations)

In regular programming practice, the way to attain the correctness is to have a complete mental model of a program and using informal reasoning make sure the thing you wrote actually matches that model. Programming skill is actually this sort of ad-hoc process where you mentally decompose a program into its constituent properties and make sure that they hold by simulating the entire thing in your head.

So to prove an algorithm correct you need to prove that its various properties hold for all inputs.

Anonymous

Anonymous Wed 08 Jan 2020 21:46:01 No.11287448 Report

Quoted By: >>11287513

>>11287257
> I took some applied math courses and some other physics courses too, but in all of them I was told that algorithms are pretty much useless for understanding, just forcing a computer to print an output is not science, you have to look for interpretations for the results using the theory, so in other words CS will never help me to give an interpretation to my results
Personally, this is what I am struggling with at the moment. I've gone through the great effort of going through the Software Foundations exercises in Coq, and my experience is the the actual proofs about the programs took extreme amount of effort, but in the end they failed provide any more understanding than the ad-hoc informal methods I had already started with. To actually do the proofs, you actually do need full understanding of the algorithms and the programs whose properties you are trying to prove to begin with.

I've begun to realize that once you take away the proofs, assuming that whatever you are looking is correct - then in fact that simplified program might be the purest expression of understanding there is. I've run into cases that convinced me that actually proving things formally actually drive me away further from understanding because it requires changing the original simple intuitive thing into something not-so simple that the PL can digest.

Also, I've found that once you go from relatively simple discrete algorithms to the continuous machine learning ones, that just by going from natural numbers and integers to rationals for example increases the difficulty factor by orders of magnitude. This is both for proving anything about these algorithms, and understanding them. So I in fact cannot do it. And neither is there any good theory in any of the papers, just sophistry. At certain point you need to reevaluate what understanding even means because it might be impossible to bring it up to the usual standard for particularly difficult things.

Anonymous

Anonymous Wed 08 Jan 2020 21:47:34 No.11287453 Report

Quoted By: >>11287843 >>11287849

>>11287257
>The way mathematics is done now makes it the enemy of reason and understanding
Basically before the age of computers the choice between constructive and classical mathematics really did not mean much in practical terms, at least from my perspective. It was just a bunch of old people arguing about philosophy, and one of the sides (the wrong one unfortunately) won because it brought more 'results'.

Today, from the perspective of modern programming, classical mathematics (based on classical logic and set theory) looks absolutely ludicrous. Type theory demonstrated quite cleanly that you can have actual programs as proofs.

And compared to classical proofs, programs have many, many good properties.

In addition to proving formal properties about them which is the gold standard of rigor, you can do things like randomized testing for example. When written in high level languages programs can be quite concise and illuminating in their structure. Most importantly, understanding is something you interact with. Programs you can change and play with. You can run them on a computer and observe their behavior. You can imagine what inputs they can get and what outputs, and test your views using a machine.

Compared to that, math is pretty much made of entirely dead letters on paper. It is all made of entirely very niche domain specific languages that are mostly useless, and that only a small number of people know. And this relentless churn of very difficult to learn, low value material is in today's world called progress. None of this is formalized, and you can only really pester the authors for insight if you somehow muster the drive to learn the material. And suggestions that the material should be made more accessible via formalization are opposed with cliques that it would take too long and interfere with 'progress'.

Anonymous

Anonymous Wed 08 Jan 2020 21:48:36 No.11287457 Report

Quoted By: >>11287513

>>11287257
(cont'd) In the early 20th century I can sort of imagine what the conditions were, so I can see how people at that time would be eager to assume that 'real' numbers exist despite cautioned arguments to the otherwise.

But now, consider neural nets and other ML algorithms. Any realistic theory of them is going to have to be based on rationals, and even this is going to be insufficient from a practical standpoint - no doubt the introduction of floats that are truncated rationals introduces its own unique dynamics to any algorithm using them. If you want good theory, you need to understand these processes. With 'reals', there is no process.

The harsh reality is that once you start playing with axioms, the computers get confused (much like humans). Once you introduce a single axiom, the whole computational basis is gone. All the proofs, all the supposed rigor - those are then all just stories, fiction, dreams. With some effort this can be made consistent, but it still falls far short of the kind of rigor that comes out naturally from simply having proofs be programs.

Mathematics is put on a pedestal while programming is considered less rigorous, but once you actually look into the behaviors of the two camps this does not actually pass muster. This present attitude of mathematicians will certainly not survive long-term. At the present time, mathematics does not deserve the prestige allotted to it.

Anonymous

Anonymous Wed 08 Jan 2020 22:05:45 No.11287513 Report

Quoted By: >>11287849

>>11287257
(cont'd) Let me just finish this by saying that I am aware that what I wrote in...

>>11287448

...where I said that rationals make things super difficult, and...

>>11287457

...you need rationals is not exactly consistent. Science is stuck between a rock and a hard place when it comes to ML, there is no way around it.

There is a certain story to be told of understanding which I do not know yet. If understanding is something you interact with, then maybe aiming to be a wizard and playing God might turn out to be the magic we've been looking for in the end.

If you look more closely, consider the close ties between ML and randomized testing and really think about what it would take to create good, robust ML models, then you will see that the requirements are identical to those for doing good science. Which might not necessarily entail proving things formally.

Anonymous

Anonymous Thu 09 Jan 2020 01:20:56 No.11287843 Report

Quoted By:

Anonymous

Anonymous Thu 09 Jan 2020 01:25:13 No.11287849 Report

Quoted By: >>11288555

>>11287453
I agree with everything except
>And this relentless churn of very difficult to learn, low value material is in today's world called progress.
I don't think one should discourage people from doing supposedly "low value" stuff.
I can see no goal to math we would have to follow.

>>11287513
I think about this quite a lot too.
Maybe there's a good way of computing with truncated rationals, namely by keeping track of the order-analysis in the programs.
What this would probably entail, however, is being aware of the number of computational steps (e.g. number multiplications) that are then performed during runtime. E.g. if you multiply 1000 times, then you can proof an upper bound of error, and also for 1000000 operations, which however will be a different one.

Anonymous

View Same Google iqdb SauceNAO 1575034660284.jpg, 297KiB, 1024x721

Anonymous Thu 09 Jan 2020 01:40:46 No.11287875 Report

Quoted By:

>>11285156

Anonymous

Anonymous Thu 09 Jan 2020 01:42:13 No.11287880 Report

Quoted By:

>>11285156
And yet, you'll still graduate and just be a code monkey that's lucky if his job isn't outsourced.

Anonymous

Anonymous Thu 09 Jan 2020 01:46:12 No.11287886 Report

Quoted By: >>11287887 >>11288833

>>11286909
There is no underlying theory its all ultimately aping information theory and logic which aren’t valid models to begin with. Mathematicians at least do not believe math is fundamentally grounded in reality, but CSfags and information theorists, logicians on occasion as well, aren’t self aware enough to discern this.

Anonymous

Anonymous Thu 09 Jan 2020 01:48:00 No.11287887 Report

Quoted By: >>11287894

>>11287886
(You)

Anonymous

Anonymous Thu 09 Jan 2020 01:55:14 No.11287894 Report

Quoted By:

>>11287887
cope

Anonymous

Anonymous Thu 09 Jan 2020 02:27:28 No.11287953 Report

Quoted By:

>>11286853
>CS is ECE for those that are bad at math.
The biggest cope of this is the typical CS major takes more math
Stay seething.

Anonymous

Anonymous Thu 09 Jan 2020 09:41:49 No.11288555 Report

Quoted By: >>11288608

>>11287849
>I don't think one should discourage people from doing supposedly "low value" stuff.
>I can see no goal to math we would have to follow.
This is true, but I cannot help but wonder if widespread adoption of type theory as the foundation of math would really increase the approachability of it. I said that formal proofs did not increase my own understanding of programs while I was doing them, but I do not know whether that would be case for programs and proofs done by others.

I think in that case, my understanding would be bolstered.

I feel that all these excuses about progress are merely just laziness resulting in gatekeeping. The cure for that is to lower the prestige of current policies one shitpost at a time until the lack of rigor in math is recognized as a widespread problem. Then you can bet people will be lining up to formalize everything. They might even begin to realize the difference between classical and constructive logic is not just one of flavor.

Anonymous

View Same Google iqdb SauceNAO B_nJ_vpXEAA_mOA.jpg, 58KiB, 360x640

Anonymous Thu 09 Jan 2020 09:43:05 No.11288556 Report

Quoted By:

you're all a bunch of 4 eyes virgins. Get laid

Anonymous

Anonymous Thu 09 Jan 2020 10:34:16 No.11288608 Report

Quoted By:

>>11288555
What's dangerous about formalization - especially at this early point in time - is the chance of getting locked into a bad language and system

Anonymous

Anonymous Thu 09 Jan 2020 10:39:05 No.11288614 Report

Quoted By:

An algorithm made this video :
https://youtu.be/znSSjrnNeSg

Anonymous

Anonymous Thu 09 Jan 2020 13:42:14 No.11288833 Report

Quoted By:

>>11287886
Shannon Harley limit has yet to be broken anon. Information theory is still very valid in telecoms

?? ???? !TdLGxo1/9U

?? ???? !TdLGxo1/9U Thu 09 Jan 2020 14:14:51 No.11288899 Report

Quoted By:

>>11285156
This is the same field I'm looking to get into eventually.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches