/sci/ - Science & Math

>>10544664
Brainlet

Normies are idiots and always gobble this shit up

Science reporting is the worst really

>>10544664
>normies
You created a thread to say this.

>>10544664
>hey guys look at me all these Normies believing this while super intellect me doesn't

>>10544680
>>10544696
>samefagging normie brainlet

>>10544664
4 independent teams assembled data from two different data sets.
All 4 teams came up with similar looking pictures.
And both data sets came up with similar shapes, which is good, since they were both looking at what we thought are black holes.
It's interesting that you alone can find some methodological error that 150 professors couldn't see. Hurry up and publish it, your Nobel is waiting.

>>10544705
>150 professors
Uhm, no sweaty, it was all her. Stop being so sexist, you incel.

>>10544705
>4 teams using algos trained on black hole sims came up with same thing
wooow

I did my PhD in Observational Astronomy and 90% of the field including my own thesis is utter bullshit

>>10544739
Interesting. Could you please share your views on why you think that is? Are there other astronomists who would agree, and have they written anything on this topic?

>>10544705
>It's interesting that you alone can find some methodological error that 150 professors couldn't see
oh they could see it. But that Nobel Prize is looking very tasty.

>>10544739
Could you expand on this?

>>10544664

>>10544664
FACT: Modern Astrophysics is less scientifically reliable than social sciences. It's both just fitting data according to your agenda, both aren't empirical research.

>>10544664
have sex, you incel freak

>>10544705
Most scientists are happy to manipulate data in order to validate their research. Also, the algorithm would probably make a similar image with data from a toilet flushing.

>>10544746

Are you retarded? Pretty much 99% of top science is bullshit nowadays. Studying something does not make you have brilliant insight into it.

>>10544846
>>10544859
>>10544866
Very Lazy Bait

>>10544872

are you even a graduate student? are you a PhD student?
there is literally no difference between an undergrad and a PhD student except a time span of 4-5 years.

Most people never get any brilliant ideas and thus produce complete and useless trash as PhD thesis.

>>10544664
This. Proof is of the second black hole, which wasn't a real image but everyone blindly believed it was. Until you show me a unedited, real photo, I don't believe anything conjured up from an algorithm and cherrypicked data to make it seem like somethings there.

>>10544872
https://retractionwatch.com/

oh no

>>10544664
>design a black hole imaging AI trained on black hole simulations
citation needed, not everyone is a plebian AI "researcher"

>>10544906
most of these are not related to astrophysics.

>>10544922
you are splitting hairs

>>10544930
fuck off retard >>10544846 >>10544859

>>10544664
How do you think imaging technology works
Youre getting triggered by normie marketing

>>10544872
There is no bait. This is not a "real Picture" like they want to make you believe, it's data fitting from random noise. You dont have to believe me, just watch the TED talk of one of the creators to see the method used:

https://www.youtube.com/watch?v=BIvezCVcsYs&feature=youtu.be

relevant part begins About 5 minutes in.

This is btw also why so many people were able to accurately predict what the "picture" will look like, because they knew what the algorithm is searching for and what is going to find will look like what it is searching for.

>>10544952

What am I supposed to be mad about now? That these images aren't "real", they're just collection of measurements assembled together?
Because that's how images normally work.

>>10545002
The issue is that the assembler program can only assemble "black holes".

>>10545006
So your confidence in artefacts which are present in the images it produces is dependant on the rate that those artefacts also arise from feeding that algorithm non-blackhole data. What's the issue?

>>10545015
>give program random noise
>produces image that looks like simulated black hole
>give program sensor data
>produces image that looks like simulated black hole
wow they look like what we expecred!

This lady is privileged as fuck, also fuck people who ask questions in lectures

>>10545015
breaking: neural network trained to identify dogs identifies dogs in white noise

desu the idea of an "image" generated this way being called a photograph is a huge stretch

"depiction" maybe

But then again they are just fighting for funding

>>10545028
>>10545046
Did you read the post you replied to?
The point is that the algorithm shouldn't be creating interesting features from random noise, and the data they did use wasn't random.

>>10545063
>The point is that the algorithm shouldn't be creating interesting features from random noise,
and yet it does

>>10544739
based

this is almost humanities level

>>10544664
So they actually didn't take a picture they just generated one through a program and the media is giving all the props to her because she is a brave jewish woman that struggled through all of the opressive white co workers.

Honestly, why would they jerk around for like 2 years, by making a shitty jpeg if it's not the real deal?

So aside from the picture, was any information gained by all this?

>>10544664
HAHA fucking this I was about to make a thread

Basically, there is nearly an infinity of possible images that fit the measurements they made with their telescope, most of which look nothing like a black hole.

So what do they do? They pick the image that most resemble what they think the black hole is supposed to look like.

They color it with the colors they think the black hole is supposed to have.

And TA-DA the first image of a black hole! Let's organize and hype simultaneous conferences around the world to share this amazing image! Science motherfuckers!

>>10544727
fucking kek

http://news.mit.edu/2016/method-image-black-holes-0606
Last paragraph.
>The authors of this paper use a highly advanced approach to learn prior knowledge,” he continues. “The application of this prior-model approach to event-horizon images is not trivial. The authors took on major effort and risk. They mathematically merge into a single optimization formulation a very different, complex sensing process and a learning-based image-prior model.

The "very different" complex sensing model was devised in the 50's
https://en.wikipedia.org/wiki/Very-long-baseline_interferometry
https://en.wikipedia.org/wiki/Closure_phase
https://en.wikipedia.org/wiki/Roger_Clifton_Jennison

The only new thing they seem to have done is feeding their simulations of black holes into their AI in order to train it to do a "better" reconstruction.
The problem is that is the most dubious part of the whole fiasco.
The construction of a better prior relies on assuming current physics is correct and that the simulations are accurate/representative.

>Assume you know what you should see
>See what you assumed you should see

The ABSOLUTE state of science.

>>10545175
>Assume you know what you should see
>See what you assumed you should see

>This first image of a black hole is thus another confirmation that Einstein's equations of general relativity are correct! Another win for science!

>now give us some more funding so we can create some more black hole pictures with our algorithm and hype it to the world so they can feel like we're advancing science so they give us funding for the rest of our career

>>10544739
based larper

>>10545175
>calling it a photograph when it's created used simulations you fed it to better represent the object you are trying to photograph
They are just playing the entire world to keep getting funding

>>10544702
Lol imagine being that faggot, no just 2 normal people calling you a moron

>>10544952
Using YouTube to validate anything automatically invalidates it.

>>10545192
I don't understand though. There really are genuinely interesting things happening in physics at the moment (albeit over a seemingly large period of time from the layman's perspective) that have been detected through legitimate means. Gravitational waves and the higgs are both much more interesting than this, even if it was a real image.

Why do they need to push this so much for >funding when it isn't even the closest thing to a 'breakthrough' of the last 5 years?

What are the odds this is invalidated in the next 35 years?

95%?

>>10545226
they know no one cares about the image

probably setting up for the girl to become a big media personality and have her own science shows since that black guy caught a #metoo accusation

>use their code and procedures and train with cat photos
>input data
>black hole now looks kitty

What's the over under on her winning the Nobel?
National Medal of Science is a lock.

>>10544664
From all this, what I'm getting is:
>AI trained
>takes seemingly random data and correlates
Are we talking about some ML bullshit? Are people really praising her for making an algorithm that literally just runs a bunch of regressions? I'm appalled, but not really surprised.

>>10545237
>Ada Lovelace
>Marie Curie
>Rosalind Franklin
>Katherine Johnson
>Katie Bouman
The titans of science. Remember these names!
(Why don't they give a shit about Noether?)

>>10545226
Gravitational waves and the higgs are not much better than this black hole photograph lol

Gravitational waves are predicted in any theory where gravity propagates at a finite velocity. Reconstructing the source of the signal they measured is akin to what they're doing here with the black hole image reconstruction, lots of fudging involved.

I'm less versed in particle physics but the same shit goes on over there and even worse, we're dealing with a collaboration of thousands of people manipulating enormous amounts of data using a bunch of intractable assumptions in the process, there is a good book on the subject https://www.amazon.com/Higgs-Fake-Particle-Physicists-Committee/dp/1492176249

A lot of modern science is absolute shit, that whole clown show will have to come crashing down at some point, when people realize they're being fed fantasy stories that don't lead anywhere

>>10545237
so did she actually have a massive role or is the media just shilling her because she's young, cute and WOMYN IN SCIENCE?
>The CREATOR of the first black hole image
but then
>Bouman HELPED develop the algorithm
Out of the hundreds who actually worked on this what makes her so significant? Surely the leaders of the research team should be getting the lions share of the recognition anyway?

YASS QUEEN SLAY!!!

incel scientists btfo once again, it's OUR turn

In the youtube talk they said they somehow mitigated the problem of training the AI to produce exactly what they were looking for but I don't really understand how.
What did they do? I understand you can take all sorts of random images, split them into patches (which by themselves do not assume any particular laws of physics) and then reconstruct from the patches, but the way they ranked the final image is still based on how much it looked like a black hole...?
Am i missing something here?