How fucked really is science? -

wtfNeedSignUp

kiwifarms.net
I know the subject of Machine Learning and it's on a constant attack by leftists and there is quite literally nothing that could be done without risking being cancelled. The attacks are basically arguing that engineers should perform social engineering themselves without seeking approval from their employers, this is done by manipulating Machine Learning algorithms to get the "right" results rather than the "true" results. So if you write a system that filters job application forms you must have the results be the same for men and women (or really, better for women), otherwise you are a part of the evil patriarchy.
You can see the results in stuff like Google's algorithm that gets progressively worse and more constrained as the years went by.

Also we can't forget that the field itself bows down constantly for authoratarian causes while using social justice as a smokescreen.
 
  • Like
Reactions: Clovis

Moral_Equivalent_of_ISIS

As bad as ISIS if not worse.
kiwifarms.net
On 4chan, here and elsewhere I've seen a lot of people taking up the mantle against science. I really don't agree with the broad claims of science being "controlled" or whatnot as a blanket statement, but on the other hand, I think people put too much weight on scientists' interpretations of their work. For example, many social science papers have shown that people with stereotypically black names (Lamar, Jamal, Tyrone) on their resume receive fewer interviews than otherwise. I don't think these studies are falsified at all (there have been quite a few of them, and this is a relatively well-researched subject). But then you often read the scientist's broad interpretation that this is the only/main cause of black unemployment, which is really hard to swallow, given the vast number of economic factors at play. A really good book about this sort of thing is Betrayers of the Truth, it comes down a little hard on the anti-science side of things, though.

On the other hand, we're typing these posts on the end-products of scientific achievement, and the 20th and 21st century have both brought such a massive wave of scientific technology that it's really hard to suggest science isn't on to something. As far as scientific self-censorship, it's easy to loudly proclaim one's devotion to pure science and investigating every possibility, but there are a lot of severely stupid people who will just insist they are correct despite all evidence to the contrary. Inherently scientists become trusted for being unbiased and independent, so they're sensitive to people trying to abuse that trust for specific political goals. Someone in this thread mentioned global warming. I think scientists became exceptionally biased on this because for years the popular opinion on the right was that global warming didn't exist and that scientists were filthy liars. Some industry-funded groups tried to pass misleading data off as science, purely to advance an agenda, which really pushes everyone else to the other side of the debate. I think this is why scientists have come to side with the opinion that massive spending and such, are required to prevent global warming.

I know the subject of Machine Learning and it's on a constant attack by leftists and there is quite literally nothing that could be done without risking being cancelled. The attacks are basically arguing that engineers should perform social engineering themselves without seeking approval from their employers, this is done by manipulating Machine Learning algorithms to get the "right" results rather than the "true" results. So if you write a system that filters job application forms you must have the results be the same for men and women (or really, better for women), otherwise you are a part of the evil patriarchy.
You can see the results in stuff like Google's algorithm that gets progressively worse and more constrained as the years went by.

Also we can't forget that the field itself bows down constantly for authoratarian causes while using social justice as a smokescreen.
Are they wrong? I know SJWs are annoying but machine learning really has its limits (I work in ML as well). It is a serious issue to use something easily capable of bias to determine stuff like people's place in society, their job, their career, or their chance of returning to prison. I'd rather machine learning's influence be slowed down a little than have potentially millions of people screwed by an overfit model.
 
  • Agree
Reactions: Dead Mime

wtfNeedSignUp

kiwifarms.net
Are they wrong? I know SJWs are annoying but machine learning really has its limits (I work in ML as well). It is a serious issue to use something easily capable of bias to determine stuff like people's place in society, their job, their career, or their chance of returning to prison. I'd rather machine learning's influence be slowed down a little than have potentially millions of people screwed by an overfit model.
Having an overfit model shouldn't be a factor. And besides the immorality of having a literal systematic prejudice built into a decision making tool because statistics are racist, imagine the situation of doing the same for something that decides whether to give parole to people in jail. You risk other people's lives because your own sensibilities. If you want to help scoeity then request more detailed data to remove factors that might cause equality (for example ask for average wage/unemployment in a person's neighborhood so that it might reduce the effects of race).
The idea that bias is something that's bad is preposterous.
 

Watermelanin

Proud self-hating degenerate
kiwifarms.net
Having an overfit model shouldn't be a factor. And besides the immorality of having a literal systematic prejudice built into a decision making tool because statistics are racist, imagine the situation of doing the same for something that decides whether to give parole to people in jail. You risk other people's lives because your own sensibilities. If you want to help scoeity then request more detailed data to remove factors that might cause equality (for example ask for average wage/unemployment in a person's neighborhood so that it might reduce the effects of race).
The idea that bias is something that's bad is preposterous.
Maybe I can offer a different perspective here:
Let's say, for the sake of argument, that race really is only skin deep and there is existing racial prejudice against black people in the legal system. A black person and a white person behaving in exactly the same manner, with the same income and background, will have different incarceration/re-incarceration rates. Now let's say you design an AI specifically to monitor these factors and, lo and behold, black people are flagged as being more likely to be repeat offenders.
Obviously, the AI isn't a racist bigot. It's just working on the data it has available. And if that data is tainted by a racist system, the AI's conclusions will just reflect the same racist attitudes. We can debate whether those initial assumptions are valid until the end of time. But it would be pretty hard to argue that the conclusion here is unreasonable if those assumptions did hold true.
 
  • Agree
Reactions: Dead Mime

wtfNeedSignUp

kiwifarms.net
Maybe I can offer a different perspective here:
Let's say, for the sake of argument, that race really is only skin deep and there is existing racial prejudice against black people in the legal system. A black person and a white person behaving in exactly the same manner, with the same income and background, will have different incarceration/re-incarceration rates. Now let's say you design an AI specifically to monitor these factors and, lo and behold, black people are flagged as being more likely to be repeat offenders.
Obviously, the AI isn't a racist bigot. It's just working on the data it has available. And if that data is tainted by a racist system, the AI's conclusions will just reflect the same racist attitudes. We can debate whether those initial assumptions are valid until the end of time. But it would be pretty hard to argue that the conclusion here is unreasonable if those assumptions did hold true.
In the specific case of re-incarceration rates there isn't any sort of provable systematic racism (or any case where the data is objective reality). But in a general case, one could always raise a complaint with his managers about the validity of the data and since most people aren't psychopaths they will likely change the task or the data components to compensate since no one wants a system that works wrong.
 

CheezzyMach

Hulkamania Brother!
kiwifarms.net
I think too many people see Science as a religion in that it's infallible. * I used to do this too *

I've said it before but Troon shit is just the modern Lobotomy/Phrenology.

Point being Science has a long history of promoting utter batshit if it's convenient and profitable to do so facts be damned.
 

Secret Asshole

Expert in things that never, ever happened
Local Moderator
True & Honest Fan
kiwifarms.net
Here are the problems from an insider:

1) Funding

This is the biggest one. Science is not cheap. Good science is REALLY not cheap. Your instruments need to be good, and a centrifuge, the most basic of scientific equipment, can cost you 8k. 'But asshole,' you might say. 'Centrifuges can be cheap!' Yeah, they can be cheap. If you don't want one that's refrigerated or can spin at high G's. Both of which are required if you want to do anything resembling actual research. Also, the cheaper the equipment, the more time your research will take because of labor you have to commit to it.

The problem is because funding is growing tighter and tighter, more daring and 'out there' research becomes less and less common, because no one is funding risky sorts of projects. This leads people to pursue 'safe' research.

Nearly all fudging of numbers is because funding is hard to come by. You have to make your initial experiments look good to get money. Many scientists are extremely desperate so they twist results so they can get money and 'prove' their hypothesis. Note, this is ESPECIALLY bad in cancer research particularly, which is going through somewhat of a crisis since funding in that area is hotly contested.

Universities typically don't fund you, I'm in a rare university that gives a lab budget, which is about ~$700 per week. This is LAUGHABLY low. If you want to fix problems with science, you have to start with the money issue. It is the number 1 problem plaguing it.

And for faggot lolbertarians, no, private interests cannot correct this. It is a problem on the national level. Not to mention private interests only compound problems. A few rich billionaires cannot fix a problem that is on the national scale.

2) Publish or Perish in Academic Research

To survive in academic research, you have to publish as much as you can. This leads to scientists splitting one experiment into multiple papers. It also leads to re-configuring or re-tooling data for multiple experiments into one. The fact of pumping out multiple articles is harming science as a whole.

3) No Money for Replication

Replication is one of the hallmarks of something solid. If something is repeatable, it is accepted. Unfortunately, replication doesn't happen often enough. Studies sit around going unreplicated because it is expensive to replicate and it usually isn't done unless someone is interested. Or something is so powerful that its goes against politics to replicate it. See Theranos.

Also cancer and psychology in particular are undergoing replication problems as in 60-70% of psychology papers can't be repeated.

4) Immortal Cell Line HeLa is wreaking massive havoc

HeLa cells, or so-called immortal cell lines have wreaked massive havoc on modern day cell lines. Its estimated that 20% of all Cell Lines are contaminated with HeLa. HeLa cells, named after Henrietta Lacks, is an extremely important cell line based off cervical cancer. However, when I say immortal, I mean it. These cells have been found floating alive in liquid nitrogen, have traveled from different buildings in the same complex on labcoats and gone on different floors of labs.

This calls into questions thousands of results of papers as the cells they've been using may have been HeLa instead of the cells they thought they were. In the end, Henrietta Lacks was exploited for her cells, so she gets to have the last laugh. Probably for eternity.

5) Failure is Unreported

Sometimes hypothesis don't work out. But that's ok, that's why we do science. However, failure is just as important as success in understanding scientific phenomenon. Unfortunately, people don't want to pay for what's good in science so failure oftentimes goes unreported and unpublished. Failing in something is a very important part. And we forgo that a lot because we need money.

7) Politics from Within and From Without

If you're looking to challenge a long-standing scientific theory, prepare for hell on Earth. Everyone will stand in your way, you'll get shouted down from he rooftops, called a cook, a crazy person and get your funding pulled. Scientific orthodoxy, once established, is INCREDIBLY difficult to challenge. Partially because a lot of research is based on that orthodoxy and because resources are scarce, people tend to get pissed if you're going against something established.

For example, ulcers aren't caused by stress typically, 70-80% of all ulcers are caused by a bacterial infection. There was an Australian scientist who challenged this. They called him an idiot, a retard, completely wrong and a waste of time. To prove them wrong, he swallowed a tube of bacteria. Next day, he was riddled with ulcers. Another example is a cycle called the Krebs Cycle. This is part of how we gain energy from respiration. Its complicated biochemistry, but lets just say people thought Krebs, the guy proposing it, was a fucking moron who was a terrible researcher. He got rejected from Nature with a nasty letter. Later on, he would go on to win the Noble prize for proving this to be correct. He framed his rejection letter in his lab.

Its this difficult to go against the prevailing orthodoxy. The stronger the theory, the heavier resistance will be. Also scientists all hold fast to extraordinary claims require extraordinary evidence. Which they do, but the arrogance surrounding this is off-putting.

There's also politics from without, as in, outside influences. People will pay for studies to have them done, governmental organizations and political movements. They will try to bend science to its will because it is viewed as the ultimate arbiter of human truth. Which I hope with this post, you'll see that it isn't.

In the end, science is run by humans. We try the best we can, it'll always be flawed, but its the best we have. And as a scientist: Always be skeptical.

I'd say the wrongfully called "hard sciences" are way less accurate than social sciences as all factual knowledge comes from reason and intuition like Mathematics, not from empirical evidence obtained through perception. See how Physics prostitutes Mathematics.
Theorical models, rather than being oversimplifying they're the most accurate representation of the world itself, which is Mathematical, and theorical models are not polluted by empirical nonsense and the delusions of perception.

But I guess my autism is too advanced for you, keep living in your normie realm with your "epic wholesome 100 facts and logic" *tips le fedora*
Reason and intuition need to be tested. You can't rely on those alone, which is where the massive fault of social sciences comes in. Miasma theory of disease and pretty much everything wrong like Lysenkoism has been 'reasoned' and 'intuitioned'. You have to test it. Without testing it, its useless. It makes sense that if I work out and get strong, my offspring will get strong too. That can be reasoned and based off intuition. Its also completely wrong.

Everything, including mathematics is colored by our perception. We do the best we can and it is never going to change because we're human beings with emotions and our views are shaped by varying experiences. Science, much like capitalism for economics, is the best system we have for exploring the world. It isn't going to be perfect. Also mathematics still relies on tangible observation, even indirect observation.

Your premises are fundamentally false. We don't have Laplace's Demon or anything close to that level of perfection. Even theoretical models have been corrupted by people and aren't isolated from perception. You're not going to get around it. Even computers programmed by humans will always be imperfect and won't be free of it.

Data will always need interpretation and people will be doing that interpretation. You just sound like a bitter mathematician. Which makes sense, since all mathematicians are bitter.
 

Ask Jeeves

True & Honest Fan
kiwifarms.net
...
4) Immortal Cell Line HeLa is wreaking massive havoc

HeLa cells, or so-called immortal cell lines have wreaked massive havoc on modern day cell lines. Its estimated that 20% of all Cell Lines are contaminated with HeLa. HeLa cells, named after Henrietta Lacks, is an extremely important cell line based off cervical cancer. However, when I say immortal, I mean it. These cells have been found floating alive in liquid nitrogen, have traveled from different buildings in the same complex on labcoats and gone on different floors of labs.

This calls into questions thousands of results of papers as the cells they've been using may have been HeLa instead of the cells they thought they were. In the end, Henrietta Lacks was exploited for her cells, so she gets to have the last laugh. Probably for eternity.

5) Failure is Unreported

Sometimes hypothesis don't work out. But that's ok, that's why we do science. However, failure is just as important as success in understanding scientific phenomenon. Unfortunately, people don't want to pay for what's good in science so failure oftentimes goes unreported and unpublished. Failing in something is a very important part. And we forgo that a lot because we need money.
...
There was inevitably a BLM article in Nature this week, it complained that using the HeLa cell line is racist and should be discontinued immediately soley on the basis of black discrimination, just absolute :lunacy: .

A journal that specializes in "failed" experiments could be interesting. It could help reduce both the publish or perish problem and the unreported failures.
 
  • Lunacy
Reactions: Pissmaster

Secret Asshole

Expert in things that never, ever happened
Local Moderator
True & Honest Fan
kiwifarms.net
There was inevitably a BLM article in Nature this week, it complained that using the HeLa cell line is racist and should be discontinued immediately soley on the basis of black discrimination, just absolute :lunacy: .

A journal that specializes in "failed" experiments could be interesting. It could help reduce both the publish or perish problem and the unreported failures.
To be fair, in the 1950s, it had nothing to do with her being black, consent was almost never sought for anyone. Also, its utterly impossible to discontinue HeLa. I mean its literally impossible because it is, again, literally immortal and everywhere. And the idea of getting paid for your cells is an extremely modern concept, one that is heavily debated. Cell lines are expensive, and royalties to people would make them even more expensive. Science couldn't function at all if this were implemented, so I'm against compensation if you donate your cells.

Its also a useless sentiment, since people have been trying to move away from HeLa as a model for a long time now. Just like people are trying to move away from mouse models. Science has been looking to try and replace model organisms and model cells for a long time now.

And that would be interesting for a journal, the problem is the money. Funding would be difficult for it.
 

Watermelanin

Proud self-hating degenerate
kiwifarms.net
In the specific case of re-incarceration rates there isn't any sort of provable systematic racism (or any case where the data is objective reality). But in a general case, one could always raise a complaint with his managers about the validity of the data and since most people aren't psychopaths they will likely change the task or the data components to compensate since no one wants a system that works wrong.
Well the problem isn't really the system itself, but the interpretation. I'll stick with the re-incarceration thing for now:
What this AI is telling us is that, of those who are incarcerated, black people are more likely to be incarcerated again at some point after release. There's nothing incorrect about that. But how you interpret this fact changes how you use this information. If it's a sign that blacks are more likely to be repeat offenders, then one should watch blacks more closely after being released. If it's a sign that blacks are already watched more closely than others after release, then we should be focusing our attention elsewhere.

I assume you're familiar with the riddle about the Airforce mechanic who's tasked with identifying where more armor is needed in their aircraft. He regularly finds fighters return to the hangar with bullet holes in the wings, fuselage, tail, etc. but never does he have to repair a fighter with a broken engine. So he decides more armor is needed around the engine, because they never make it back.
 

betterbullocks

deep anal and cool ranch doritos, shes got it all
kiwifarms.net
Pop sci articles are, always have been, and will continue to be, prone to errors and penny chasing clickbait. If you actually immerse yourself in the legitimate scholarly scientific literature around a field, things will brighten up a bit. Not entirely, and the degree to which this is true varies field to field, but it will.

I'm a physics major and nothing makes my spine tingle with cringe like seeing "tHe CaT iS dEaD aNd AlIve?!?1" but reading through some peer-reviewed papers on a spacetime topology can be equally as novel as reading about time travel for the first time, albeit a lot more dry and a bit less sexy. Still, if you find enjoyment in it, and you can get utility out of it, pursue it. It's worth your time.

It also helps to get a solid grounding for the dialogue on a topic before you read papers. A lot of social or health science papers might seem redundant, like it's all people rehashing roughly the same shit and coming up with wildly different results. But if you get a sense of the conversation up to that point, you can better understand the purpose behind specific experiments, studies, and reviews. Beyond the literaly purpose of the article, what questions are they REALLY asking?

Not that the layman could possibly give two shits about making all that effort just to understand why the permafrost is thawing and if anyone should care
 
Tags
None