Black Hole Image Gets an AI Upgrade!
VloĆŸit
- Äas pĆidĂĄn 16. 05. 2024
- Get Surfshark VPN at surfshark.deals/launchpad - Enter promo code LAUNCHPAD for 3 extra months for free!
00:00 Introduction
00:46 The new PRIMO image
04:06 Surfshark VPN
05:20 How PRIMO's AI works
10:02 Will PRIMO image Sag A*?
10:27 Next Generation EHT!
11:20 Thank you Patrons!
đ Subscribe for more: czcams.com/users/christianread...
đ Share this video with a fellow space traveler: âą Black Hole Image Gets ...
đŽ Watch my most recent upload: goo.gl/QbRcE2
đ Help me improve the channel by joining the community on Patreon
/ launchpadastro
đ Check out Launch Pad merchandise!
teespring.com/stores/launchpa...
Disclaimer: Some of these links go to one of my websites and some are affiliate links where I'll earn a small commission if you make a purchase at no additional cost to you.
đ§ References:
Medeiros et al. 2023 "The Image of the M87 Black Hole Reconstructed with PRIMO": iopscience.iop.org/article/10...
Medeiros et al. 2023 "Principal Component Analysis as a Tool for Characterizing Black Hole Images and Variability": iopscience.iop.org/article/10...
â Let's connect:
For business inquiries - chris AT christianready DOT com
Twitter - @launchpadastro
Instagram - @launchpadastro
Facebook - / launchpadastronomy
Discord - / discord
đ c/o Christian Ready
P.O. Box 66
Westminster, MD 21158
United States
Earth - VÄda a technologie
I still find it astounding we managed to use the entire planet as a telescope, science is just awesome.
Yep, some clever apes we have here :)
@@LaunchPadAstronomy I would not call some clever that does not understand the basic of dark star science.
You are not clever and i would suggest not an ape.
Wait until they build telescopes the size of the solar system by using fleets of satellites orbiting the Sun!
@@peterjackson2666 Oh Boy, I do wish I could be alive to see that! But the people who are here, will be in for a treat! I can just imagine millions of telescopes, all orbiting in the Kuiper Belt, I pray that mankind does, in fact, continue to reach for the heavens!
@@peterjackson2666Why stop there? Build mega telescopes across the entire galaxy so we can directly image exoplanets in andromeda
".. I tried.... and they stopped me."
That's a Supervillain origin story if ever I heard one.
Inquiring minds want to know more. He canât just say something like that and not elaborate!
@@dogcarman đđđđFully Agree,lol
Itâs like they gave the calculation a pair of glasses đł
Yeah, the original was like looking through chocolate rain.
â@@lifeisstr4nge bro, leave the guy alone.
Thats cool to see you commenting here :)
So if I got it right... this isn't an upscaled AI image, but what's happening here instead is the original EHT image was totally theory agnostic in order to avoid bias, and now that we know that it is in fact a black hole ring shaped by general relativity PRIMO can apply known physics to the original data to resolve the image in accordance with known theory... I think? Thanks as always for the great videos!
Yes, that's exactly right! And thank you very much!
đ that deadpan âI tried, and they stopped me.â Fantastic & informative video, as always!
Love the demonstration with a compact disk.
Thanks!
The only thing that boggles my mind more than these new images of M87 is the fact that Christian, who has the best Astronomy/Astrophysics videos, has 217K subscribers instead of 217M !
I'd
Maybe he should consider calling himself Christian Kardashian..
Sometimes the number does not matter, the quality of audience is more important. I'll rather prefer those noisy stray people to stay out of the comments space.
đđđđ,Thats Great,lol
Woow thats amazing, seeing capabilities of ai even in cosmology is exciting.
Your obvious excitement is soooo contagious. â€
I like how the new image shows a slight clockwise spiral pattern indicating the spin.
Even though you continue to make videos, please donât get de-motivated on making your videos, the amount of information provided in such a good quality video is awesome, PLEASE KEEP IT UP MAN! đ
Thanks! No demovitation here!
@@LaunchPadAstronomy Awesome dude!
Thanks Christian, you've answered some other questions I had about the original image, all that I can do is smile, smile, smile!
it's like running a magical "enhance" filter on a blurry 3-pixel license plate photograph. except you pretend that you know the license plate number.
Hope they will succeed with ngEHT. An actual stop motion of the black hole... wow. Can't wait!
Those last animations look terrifying. They give me the impression that someone pulled the drain plug on the universe.
Interesting explanation of what weâre actually seeing. I assumed we were looking at the accretion disk surrounding the event horizon.
1:34 Christian I had no idea you were a mastermind of some kind
Itâs thinner now because M87 decided it had better slim down a little if people are going to be taking photos of it and publishing them on the internet.
Those images are incredibles !
But most important, YOU are incredible, Christian.
You are so kind, thank you so much!
It seems odd that the brighter ring areas don't match the same positions as the original EHT 2019 composition. The bright area on the left at ~255° is gone and the lower mid-left is brighter with a dim space where it was brightest. It seems odd for the same data to present such different results. OK, it's rotating. Will Webb be imaging this black hole? And why should a black hole look like it's wearing a hat from some angles? The 90° bend is anti-intuitive. Thanks. It's been a while. I hope you've been well.
Thank You for material.
Thanks! đ
Any (space)time!
I really like your new graphics. Good job!
Thanks!
Hey Christian i wanna know why the heck there is a massive gap between stellar mass and supermassive black holes?
So do I! :) But more seriously, I have some videos that talk about evidence for intermediate mass black holes. Here's one of them: czcams.com/video/VcPMLDu7BTs/video.html
3:35 why would a doppler shift increase the brightness? Wouldn't that decrease the wavelength?
Good question. My guess is: the increased doppler shift frequency from the region of the accretion disc moving towards us, results in more energy incident on the antenna than the region that moves away from us ( since E = hf ).
It is not technically correct to say that each component antenna combined to form the image is seeing a different part of the image. Each of the individual antennas see the whole image, just at poor resolution. Effectively, each antenna has the "whole image" within a single pixel.
As you add together that single pixel of different, separated antennas, you make the resolution of the resultant pixel much finer. By changing the relative phase before adding them up, you shift the location of that finer, resultant pixel (constructive interference occurs for a different pointing angle). Doing a lot of different phase adjustments you to form a lot of different pixels of the final image.
The problem is, if two antennas are spaced more than half a wavelength apart, then there are multiple direction with that peak gain. The farther separated they are, the more different directions have that same gain (called grating lobes in passed array radars). Adding more antennas to combine reduces the number of grating lobes (and increases sensitivity)
When adding up different antennas using interferometry, you are working with the phase at each antenna. If the phase difference is 10°, you don't know if it is 10, 370, 730, ... degrees. This means when you are adding up the sources with phase adjustments so the as up constructively in the desired pointing angle, they also add up constructively in other directions (called grating lobes in passed array radars). What you get is not just what is at the pointing angle, but the sum of that and all those grating lobes (plus other stuff, just at lower gains). The more wavelengths apart your individual antennas are, the more grating lobes you have to contend with. There is guesswork in combining a sparse interferometer whether it is the scientists doing it or an AI.
Exciting stuff!
I want to know more about Magnetohydrodynamics now.
Alright who fatshamed the black hole
đđđ
Angry Fat Karen Blackhole đ
đđđ
Amazing Images Christian,And Another Great Video,Ive Been Super Busy,But Realy Do Love Your Videoâs..Stay Safe,and God Blessâ€ïžđđ»âšđđ,I Have 6 Scopes Now..Im Doing My Partđ
I knew Principal Component Analysis I learned about decades ago will be useful!
Just curious, did they use linear or non-linear (kernel) component analysis?
Not sure, but I'm guessing kernel. You can check out the papers referenced in the description and they'll tell you all about how they did it :)
LOL @ "...I should know... I tried... and they stopped me... đ€š"
The new image is cool, but itâs also slightly unnerving, the center just looks so... black
Also the knowledge itâs a black hole
I know what you mean.
I can only imagine your budget to cove the earth in dishes.
It may have been a good thing they stopped you. LOL.
This channel reeks of professionalism of the highest degree
Thank you so much, I can't tell you how much I appreciate that.
@@LaunchPadAstronomy I love what you do and how you do it. If my appreciation keeps you going - I'm honored and glad. Keep up the great work :)
Is that what that is? I thought it was Hai Karate..
(1) The radio waves that we can't see, gives the data; (2) the visual image is a radio image that we can't see; (3) EHT synthesizes an effective antenna aperture of Earth's size; A.I learns about 30K simulated models of black holes and refines the image! I'm more and more convinced that REALITY is not absolute and is in the eyes of the beholder.
Great video⊠congrats on 1M subs⊠almost 1M!!! đđđ
Almost there, just 800K to go!
1:44 lol
Japan article says black hole image is biased. Petabytes of data with only megabytes of image? something fishy here.
Imagine if there was a telescope on the moon? đđđ
Yeah, but most discussions of radio telescopes on the Moon talk about being on the far side. That decreases how much that improves the interferometric baseline. That being said, it would still be a big improvement if the Moon data is collected from much of an angle. The resolution of the Moon data decreases with increased scanning angle (off of perpendicular to the antenna), proportional to cosine of the scanning angle. However the Moon array can be quite large, so resolution is still good. The interferometric baseline has a factor of sine of the scanning angle, but it is the multiplied by the distance to the Moon. Since the distance to the Moon is almost 30 times the Earth diameter, that adds a lot. The problem is, the larger the gaps in the interferometric baseline, the more grating lobes there are. This means other directions with just as much gain as the desired pointing direction.
I was just listening and at the end I thought i am watching Arvin Ash :D:D stay curious my friend
He's my twin brother. Happens all the time.
There's some light data where the black hole should be. Is that an antumbra?
Probably not but the physical aspects of what is represented are still being determined.
This looks like the self eating serpent eating its own tail positioned in a âïž
The AI can paint anything you want on there.
Yes but this is more like a calculation. We feed it the information and it gives us the closest result
What, so the new image is blurred? The unglued version is way better
These images literally scare me
Q ???
Looks like Q-Anon has a new logo.
i'll be much more excited when there's no "AI" in the name and there's just more data instead. sure, you can feed a blurry image of a cat into midjourney and ask it to create a cat image, and it will happily create a great image of a cat. just not the cat you took the blurry photo of.
Daniel Gonzalez said it well: "..this isn't an upscaled AI image, but what's happening here instead is the original EHT image was totally theory agnostic in order to avoid bias, and now that we know that it is in fact a black hole ring shaped by general relativity PRIMO can apply known physics to the original data to resolve the image in accordance with known theory."
Just the fact that the data allows so many different interpretations doesn't make it very trustworthy. Maybe there wasn't any black hole to begin with and it's all in the interpretation. I mean if they somehow add many more sampling points and improve their SNR to the point where there will be no room for interpretation, it will be a "picture" of a black hole. But so far the data is very unreliable at best.
You gonna sell so many surf shark VPNs to those Utah Mormons since they can't freely watch porn anymore lol
Try harder next time and hold the Earth ransom for $1,000,000 this time.
đ€đ«đ€
Why are the two "images" so different (i.e. where the brightest areas are)? It's hard to have faith in either of them because of that.
Itâs partly because the PRIMO version wasnât required to be model agnostic, but also because there is still work being done to assess the physical properties of the image.
@@LaunchPadAstronomy Thanks. So the PRIMO version used an assumed model to construct the image? You can basically come up with anything you want doing that. The first version has a bright spot at 9 o'clock. The second doesn't. The first version is dim at 7 o'clock while the second version has it brightest there. Honestly, I don't accept either of them as representing reality at this point.
@@JCO2002 Well they're not exactly what it would look like, but they _are_ our best guess at what it might look like.
@@iveharzing Best guess at what it might look like... I can guess how many angels can dance on the head of a pin. I suspect there's good data there, but the "publish or perish" prerogative and the current trend toward click-bait science (Avi Loeb et al) has hyped it somewhat in to the realm of fiction. But, what do I know...
đź l see first
Stop AI
scientists kneeling and worshipping at the alter of AI
This is right out of the movie's
You're being simple.
Facts I mean the original picture from 2019 is a Ai photo based on data which like all data isn't 100% perfect, scientists believe this is an achievement but all they have done is make ai photos, we will never get a real photo of a blackhole
How is this an upgrade? At least the original contained authentic data and not something generated by an AI
This one contains exactly the same authentic data as the first one.
It just uses newer and better algorithms taking into account confirmed models.
So actually more data for smaller margins of error.
Also, the original also used algorithms and less advanced AI- not to mention it's an average of dozens of estimations!
@@andyd4292 hence it being blurry. the blurriness is largly the result of AI uncertainties. the new images has less of those. In fact they added in some uncertaintiy blurriness just in case. The reconstruction algorithsm was actually a lot more certain and precise than this image.
Both images were generated by "AI", based on the same data. However, PRIMO didn't have to be model-agnostic this time. I discuss the differences in the video, but let me know if it didn't make sense and I'll be happy to clarify.
@@LaunchPadAstronomy Thanks for the clarification. I posted before watching, since I had already heard the news.
Im sick of these highly processed images, false colors, now a damn AI massaged pic. Nothing is real.
I can relate to that. Me for myself I think they are trying their best to reveal what is extremely difficult to make visible. But you have to trust the whole path and theory behind it âŠ
The data is real. Since most of the light emitted is not in the visible spectrum (which is only a tiny fraction of the light spectrum) the data has to be processed into something we humans can see. That is a limitation of our eyes, not the data received from the telescopes.
Facts, I mean we have given an computer made image basically an ai image based on data which like all data isn't 100% reliable and ai makeover, we have just wasted money on ai images because we can't get a real picture of a blackhole
This is not science, it's virtual reality.
What do you mean? Do you not know what science is?
Garbage in, garbage out
@@nobody7158 thatâs a resounding no from the special needs student
You say that features may be artifacts and then you say that the new picture is a more accurate representation -- those two statements are inconsistent.
Regardless, I think you are downplaying how much this picture may be a figment of the algorithms imagination. The algorithm is "trained" by models -- what if the models are crap?
- - -
Bottom line: At this time, the reconstructions aren't falsifiable -- this is pseudo-science.
That's why the first images were done by teams that were separate so each was agnostic. Once they had an image they were able to refine things. It's not pseudo science.
So.... it's fake.
Not fake at all. Rather, it's reconstructed.