|
me and a neuroscience friend spent a good hour giggling together about how computer scientists are idiots when it comes to thinking cs applies in even the smallest degree to any other field and how programmers are be default, when talking about non-programming, ridiculously wrong knowitalls
|
#
?
May 18, 2022 03:02
|
|
|
|
| # ? May 4, 2024 00:02 |
|
|
tbh it's not computer scientists who are that way, it's people with a bachelor's or _maybe_ an MS in computer science. generic computer touchers trying to pass themselves off as "scientists" of any stripe is absolutely symptomatic of what you're describing, though
|
|
#
?
May 18, 2022 03:04
|
|
|
programmers as synonym to computer scientists was basically finished by the late 60s, yeah
|
|
#
?
May 18, 2022 03:09
|
|
|
Achmed Jones posted:tbh it's not computer scientists who are that way, it's people with a bachelor's or _maybe_ an MS in computer science. generic computer touchers trying to pass themselves off as "scientists" of any stripe is absolutely symptomatic of what you're describing, though Nope, most computer scientists or programmers when they start talking about anything that isn't to do with computer are utterly batshit wrong, because they don't understand how deep and complex problem spaces outside of CS are. So you get programmers pontificating about sociology or about brain to machine interfaces without realising that this stuff has 99.999999% of the time has already been studied, and is either a) explained in an undergraduate textbook and all their guesses are totally off the mark, or b) is a MUCH more complex problem than it appears and we don't even know enough to know how deep the rabbit hole goes
|
|
#
?
May 18, 2022 03:11
|
|
|
No I'm pretty sure that's wrong
|
|
#
?
May 18, 2022 03:12
|
|
|
It's a symptom of the inherent god complex that comes with being able to pick up a reasonably full understanding of a field within 4 years, and the fact that computers have been created by humans, and there's literally no complexity analogue for stuff like cell microbiology, chemical signalling, or or even stuff like "how people behave when walking down the street" in computers.
|
|
#
?
May 18, 2022 03:14
|
|
|
all the bci groups i know of are half neuro peeps and half hci and general cs peeps. they do a lot more optogenetics than they used to but the current state of the art is like one fuckin baud and signal processing poo poo will have to go into improving that in addition to neural poo poo, come on i really do think you mean programmers as opposed to cs peeps
|
|
#
?
May 18, 2022 03:15
|
|
|
alexandriao posted:It's a symptom of the inherent god complex that comes with being able to pick up a reasonably full understanding of a field within 4 years, and the fact that computers have been created by humans, and there's literally no complexity analogue for stuff like cell microbiology, chemical signalling, or or even stuff like "how people behave when walking down the street" in computers. you're reiterating what i'm saying but seem to think that you're arguing. the people who have been doing it for four years are not computer scientists.
|
|
#
?
May 18, 2022 03:17
|
|
|
bob dobbs is dead posted:all the bci groups i know of are half neuro peeps and half hci and general cs peeps. they do a lot more optogenetics than they used to but the current state of the art is like one fuckin baud and signal processing poo poo will have to go into improving that in addition to neural poo poo, come on The programmers I've known had a smidgen of sense to realise that they were middle class, so all their (wildly incorrect) pontificating about sociology and economics came with that caveat, less so for bci stuff. The computer scientists I've known didn't even bother going on google scholar to confirm their theories about human social behaviour before writing 10 paragraphs based on an axiom that's literally stated as a common pitfall in a socio 101 book lmao
|
|
#
?
May 18, 2022 03:17
|
|
|
alexandriao posted:It's a symptom of the inherent god complex that comes with being able to pick up a reasonably full understanding of a field within 4 years, and the fact that computers have been created by humans, and there's literally no complexity analogue for stuff like cell microbiology, chemical signalling, or or even stuff like "how people behave when walking down the street" in computers. I’ve been doing this for 6 years and the only understanding I have of anything is android, iOS, and Java/kotlin services. when I graduated from uni I knew fuckall and I didn’t start to fully understand things until my second year in my job
|
|
#
?
May 18, 2022 03:18
|
|
|
I'm stating that any attempts to differentiate these groups in this case will meet with failure imo because it's based on "having knowledge of a single domain and not understanding complexity of non-computer domains leading to arrogance" which both CS folk and programmers have in truck loads
|
|
#
?
May 18, 2022 03:19
|
|
|
sorry i thought this was the "complain about computer people" subforum. can you link me to the right one
|
|
#
?
May 18, 2022 03:20
|
|
|
and i'm saying i think you're full of it when it comes to computer scientists, ie people doing CS research (you're clearly right about generic computer touchers, but this hasn't been news to anybody for a solid decade+ at this point)
|
|
#
?
May 18, 2022 03:21
|
|
|
alexandriao posted:It's a symptom of the inherent god complex that comes with being able to pick up a reasonably full understanding of a field within 4 years, and the fact that computers have been created by humans, and there's literally no complexity analogue for stuff like cell microbiology, chemical signalling, or or even stuff like "how people behave when walking down the street" in computers. Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain.
|
|
#
?
May 18, 2022 03:21
|
|
|
quarantinethepast posted:Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain. lmao
|
|
#
?
May 18, 2022 03:21
|
|
|
i used to have opinions about application architecture and design patterns and stuff but then i started to do security stuff as my job title and now i no longer give a poo poo. it owns.
|
|
#
?
May 18, 2022 03:22
|
|
|
quarantinethepast posted:Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain. compsci yes, cog sci yes, neuroscience no, lol the original intention via rumelhart was to link constraint satisfaction problems to them. neocognitron (convnet) was the only biologically inspired one, and that brutally marginal. actually the main inspiration was condensed matter physics, thats from the harmonium the grand fundamental link of cs to physics is in qc and in condensed matter structure of second order phase transition in satisfiability. sat is a physical lattice, like ferromagnetic matter is (phase transition like nonferromagnet => ferromagnet, not like water => ice. extensive thingy in phase transition is difficulty). thats a fundamental if obscure grounding to the real world. deolalikar tried to prove p!=np from it and failed cuz proof goes through for xorsat fine but xorsat is in p bob dobbs is dead fucked around with this message at 03:27 on May 18, 2022 |
|
#
?
May 18, 2022 03:22
|
|
|
programmming isn't a science
|
|
#
?
May 18, 2022 03:24
|
|
|
quarantinethepast posted:Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain. https://youtu.be/-15VC4Yxzys
|
|
#
?
May 18, 2022 03:25
|
|
|
quarantinethepast posted:Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain. they also do a good job of modeling belief acquisition and provide support to coherentist views of diachronic justification and stuff. they're cool. i have no idea how they map to physical structures but im gonna guess it's not super close it's funny cause i have absolutely no idea how to use any ml libraries to categorize pictures or wtfever, but i've written neural net poo poo that (at least used to) sounds impressive on a resume or at a dinner party. even though it was just bog standard "traverse the thing, do do a math, stop when change between iterations is close enough to zero"
|
|
#
?
May 18, 2022 03:25
|
|
|
alexandriao posted:sorry i thought this was the "complain about computer people" subforum. can you link me to the right one link
|
|
#
?
May 18, 2022 03:25
|
|
|
bdid, ya my poo poo was constraint satisfaction straight up
|
|
#
?
May 18, 2022 03:27
|
|
|
science is mostly sitting around not really understanding poo poo for hundreds to thousands of years than some fucker comes up with some poo poo and the rest falls into place. the fucker with some poo poo was had in physics in 1600s, general darwinian evo biology in 1800s, cs in mid 20th century, neuroscience not yet. but its that relation w the fucker and their some poo poo that determines understanding and productivity not artificiality but in turn the some poo poo is pretty relative to your vision. in the span of all physics the fucker w some poo poo came around 1600s, but in high speeds and small scales its 1900s. with respect to wibbly chaotic bits its really not yet. etc. vast portions of the edifice of 'computer science' just dont have their fucker yet. consider the navier-stokes problem, which ended up getting shoved from math proper to cs by the reformulation of disciplines: still no fucker bob dobbs is dead fucked around with this message at 03:35 on May 18, 2022 |
|
#
?
May 18, 2022 03:31
|
|
|
quarantinethepast posted:Neural networks are a good representation of biological nerve systems and brains, and are used outside of compsci in cognitive science. Although they by no means currently capture the full complexity of a human brain. It would have been a good idea to do a basic Google search before making my claims: https://towardsdatascience.com/heres-why-we-may-need-to-rethink-artificial-neural-networks-c7492f51b7bc Got got I have
|
|
#
?
May 18, 2022 03:37
|
|
|
yeah, see? programmer right there. they were shittin on rumelharts thing in the late 1980s for being ridiculous biologically but the fact remained that the thing worked and the other putative computational models sucked. they still suck, and every few years neuro peeps come out with another one that sucks. rumelhart before he died terribly ironically was always open about backpropagation bein fuckin lol, according to my old advisor, but it did work they didnt even use the term neural net back then. "multilayer perceptron", which you never fuckin hear anymore
|
|
#
?
May 18, 2022 03:39
|
|
|
schmidhuber also loves to poo poo up the mailing lists yelling about how its all just reverse mode ad. which isnt false, lol. but the reverse mode ad peeps didnt know jack about csp's, is the rub one of the things rumelhart et al published right alongside backprop in the 80s was the boltzmann machine, which is really really really just-a-lattice-with-poo poo-on-top. hintons still wailing at that occasionally to this day bob dobbs is dead fucked around with this message at 03:48 on May 18, 2022 |
|
#
?
May 18, 2022 03:45
|
|
|
quarantinethepast posted:It would have been a good idea to do a basic Google search before making my claims: Yep, the more we learn about neurons the more complex they get. You don't just have the myriad connections themselves that create higher layer signalling properties, but also the connection type, the delay, the intensity, etc. but there is also chemical and protein signalling in addition that ups the complexity tenfold. It took Folding@Home... how long? to replicate all the different ways a single protein can fold. It's months to years IIRC. And then on top of that, individual cells can have their own behaviour and "memory" and "environmental knowledge". Plus the tools we have for neuroscience study are extremely blunt instruments. FMRI can't be compared across different subjects or even across multiple uses of the machine because of blood flow and how it slices things up. With the other non-invasive methods you can have either intensity of signalling, or specificity of signalling. You either can get very very detailed information about millions of brain cells at once, or you can get not very detailed information about only a couple of hundred thousand. Invasive methods you have to rely on someone to get brain damage and to already have had probes inserted into the area you want to study, and you have to make the test suitable to be run on a laptop (because they are most likely on the other side of the planet), and you have to get their consent. And there are just so so many ways that can gently caress up.
|
|
#
?
May 18, 2022 03:47
|
|
|
quarantinethepast posted:It would have been a good idea to do a basic Google search before making my claims: Then read that article further before linking it. Because it contains genuinely funny meta-irony.
|
|
#
?
May 18, 2022 03:51
|
|
|
optogenetics is gonna be it, prolly folding proteins is a particularly poor example of a thing unconstrained by cs poo poo because its just ('just') an absolute fucker of a constraint satisfaction problem, iff the protein only has the usual amino acids without more bullshit (and there really is some more bullshit a huge portion of the time). its actually unusually amenable to computational methods
|
|
#
?
May 18, 2022 03:51
|
|
|
bob dobbs is dead posted:optogenetics is gonna be it, prolly Oh hey that has a solid answer. It's "no".
|
|
#
?
May 18, 2022 03:53
|
|
|
Every loving field, buncha twerps staring at the clouds, describing the animals they see.
|
|
#
?
May 18, 2022 03:55
|
|
|
bob dobbs is dead posted:optogenetics is gonna be it, prolly That still requires cracking open the skull though 😬 But yeah you're right, protein folding is one of the few low hanging fruit I have for this though because most of the other things that come to mind we haven't even attempted to simulate because of the complexity. I would say tho that protein folding being amenable to computers and still being an utter gently caress of a computational problem to solve kinda proves my point even though it isn't an ideal example in the space
|
|
#
?
May 18, 2022 03:56
|
|
|
its got the aesthetics, if it looks like mad science mad science poo poo will get done
|
|
#
?
May 18, 2022 03:57
|
|
|
Kazinsal posted:it's a joke about hilbo being a teenager yes but im 19, two years above when libertarians would be interested
|
|
#
?
May 18, 2022 04:29
|
|
|
when it's a teenager it's actually called neoliberalism
|
|
#
?
May 18, 2022 04:34
|
|
|
the pro side of this is that computer touchers do know how to touch computers, so when they enter into a legitimate collaboration with professionals and academics from other fields who know all about those fields and have some computational needs, it can be really beneficial for everyone. but ofc nobody benefits from cs majors reading twitter all day and Forming Opinions
|
|
#
?
May 18, 2022 04:49
|
|
|
you'd be surprised at how little personal benefit I've managed to get out of having a science education in the world of computing. No one wants to pay any money to make scientific software any good, it's all made by grad students who need it to work or they fail and sad people at perkin elmer
|
|
#
?
May 18, 2022 04:52
|
|
|
yeah but if that grad student teams up with a cs grad student, that's two phds from one project, and the project will turn out better
|
|
#
?
May 18, 2022 04:54
|
|
|
alternatively the project will be hijacked to showcase the cs grad student's advisor's pet language/data structure/algorithm/software engineering technique. because why should anything be good
|
|
#
?
May 18, 2022 04:57
|
|
|
|
| # ? May 4, 2024 00:02 |
|
|
DELETE CASCADE posted:alternatively the project will be hijacked to showcase the cs grad student's advisor's pet language/data structure/algorithm/software engineering technique. because why should anything be good Or the cs grad student leaves halfway through to join Google
|
|
#
?
May 18, 2022 05:00
|
|