That's the kind of thing where I can't even imagine how the project conceivers and directors can even have the creativity and chutzpah to envision a $300 million dollar lab made by boring dozens of tubes 2 kilometers into the south pole to create a giant cubic kilometer of sensors.
There are precedents for this. ANTARES and DUMAND both used similar approaches although they used strings in the Mediterranean and Pacific. I suppose using ice makes things easier in that you don't have to worry about sealife and ocean currents but the logistics are going to be a lot tougher and there's probably more ambient light to worry about.
> The incredible clarity of the Antarctic glacier, revealed by the construction and operation of IceCube, will allow the spacing between light sensors to exceed 250 meters, instead of the current 125 meters in IceCube. The deployment of sensors on strings with larger spacings will enable the IceCube-Gen2 instrumented volume to rapidly grow at modest costs.
I worked with a former SW technical coordinator who came right from IceCube. The software that runs their analysis and stuff like that is insane. Many, many terabytes of data to crunch. It's very cool stuff
I did an internship with the Amanda (Antarctic Muon and Neutrino Detector Array) project - IceCubes "Pathfinder" - in the mid-1990s, and found it one of the most rewarding experiences in my life. There are over 30 years of work from hundreds of people in this project, and to see how it has really made it from "madcap folly" to "a new lens to the universe" gives me the shivers. Hat's off to them.
I went to UW Madison and now work there and know someone who did a rotation at IceCube (doing tech support). It would be a really neat life experience but Wisconsin is already too cold for my taste.
What they do it totally out of my field, but I love hearing about it. It's a big source of pride in the university.
Using neutral networks to pre-filter data, especially networks trained on stimulations, should be seen as somewhat suspect as it potentially introduces bias into the analysis that is very hard to quantity, let alone control.
Unfortunately, the current hype around neural networks incentivizes scientists to use them even in situations where there is no need. (maybe in this case there's no other way, I'm not sure...)
Neural networks for data selection in particle physics have been used since before the current AI hype cycle took off. This means there are already some established guidelines on how to mitigate and conservatively estimate pontential biases. I certainly hope the IceCube researchers have followed them.
I promise you, in particle physics it is seen as somewhat suspect. But a very careful analysis with machine learning costs a couple PhD-years: it would be even more suspect to forgo an improvement on a 400m dollar experiment.
As you say, it's easy for a neural network to pick up on simulation artifacts, rather than real physical features. The appendix of the original paper [1] explains how they quantify the failure modes of this approach. One kind of cool approach was training an adversarial network to corrupt simulated backgrounds to make them signal-like. The details are sparse, but it sounds like fooling the classifier required a level of corruption that would have been noticeable when comparing the simulation to data.
Can you do this wrong? Sure, but you can also go wrong with the feature engineering that particle physicists have used in cases where they don't trust machine learning. You won't really know if the methodology is sound unless you:
- read the paper
- follow some references (the paper is a short-form "Physical Review Letter", which means it's impossible to give any meaningful review using the text itself)
- hopefully find some more detailed description of what they did
The last point is often the difficult for particle physics collaborations, they are generally a bit protective of their data and internal working.
So a volumetric camera of sorts utilizing existing clear ice as a transparent rig and a datacenter on the south pole. Probably deserves more recognition than the rapper it they stole the name from lol
The most impressive part to me (having wintered for IceCube) is how well organised the stack is, and how reliable the in-ice hardware is.
The systems folks have had a decade to iron out most of the bugs and there are fairly tightly scoped teams that work on different components. The failure rate of the DOMs is extremely low for something that's deep in the ice and impossible to physically access. They were very well built and designed, but it helps having people on call 24/7.
All the telescopes here have some sort of data center. Icecube happens to have a lot more computers because of the sheer number of detectors, and a desire to filter data so that preliminary/candidate detections can be sent North as soon as possible. Though unlike the other installations, IceCube is far more sensitive to downtime.
The Event Horizon Telescope setup at SPT probably takes the crown for superlatives in terms of data rate, clock accuracy, etc.
Our ability to extract such a small signal from noise that is so large we don't even have words for it is truly extraordinary.
It strains credulity to the absolute maximum -- and yet these results are repeated, again and again, all over the world by different groups, using different equipment, based on different detection principles, using different methods.
My first thought was that this person is a grifter, but then I remembered Hanlon's Razor: "Never attribute to malice that which is adequately explained by stupidity."
Every single time this science hits headlines I have at least (5) seconds of trying to understand how O'Shea Jackson suddenly had a scientific addendum to neutrinos and if that is as-cool or cooler than "No Vaseline".
Straight outta Amundsen-Scott / Crazy observatory named Ice Cube / From the gang called "physicists with attitude" / when I'm called on / I find a lepton / they emit Cherenkov radiation
There’s a lot of other rappers that have made the move from rap into tech. Chamillionaire went from writing Ridin’ Dirty to investing in Lyft. Even wears glasses now.
Probably! There were lots of us grad students on Francis's floor! But I don't think he cared very much about that sort of thing. (Maybe more than he cared about being on my committee, though? OK, low bar.)
I really do think it's named that because it looks like a cube and it's made in the ice. So, y'know, icecube.
About 12 years ago someone told me Francis said they would've named IceCube something different if they'd known about the rapper. I can't find any references to that quote now, though.
Absolutely yes, though the original naming committee (Francis?) probably didn't think of it - it's literally a cube of instrumented ice. Not so much the other way around I suspect. A while back some IceCube folks paid for a shout-out from Ice Cube, the rapper (I guess they used Cameo).
Yeah but the question was about how familiar the people involved might be with the rapper. I would imagine any 30+ American or European (probably) would.
He was a founding member of N.W.A, which was very influential to the rap scene:
>They were among the earliest and most significant popularizers and controversial figures of the gangsta rap subgenre, and the group is widely considered one of the greatest and most influential groups in the history of hip hop music
>Rolling Stone ranked N.W.A at number 83 on its list of the "100 Greatest Artists of All Time".
I was thinking about how it was good he observed potential tau neutrinos and not some motherf*ckers in a blue uniform, otherwise this could have gone different. He could have swarmed!
Incredible to image a square kilometer of pure ice transformed into a sensor to detect rare subatomic particles!
Obviously I'm glad such people exist.
https://icecube.wisc.edu/science/beyond/
> IceCube-Gen2, a ten-cubic-kilometer detector
> The incredible clarity of the Antarctic glacier, revealed by the construction and operation of IceCube, will allow the spacing between light sensors to exceed 250 meters, instead of the current 125 meters in IceCube. The deployment of sensors on strings with larger spacings will enable the IceCube-Gen2 instrumented volume to rapidly grow at modest costs.
<https://www.wired.com/story/neutrino-detectors-could-be-used...>
https://en.wikipedia.org/wiki/Pierre_Auger_Observatory
https://www.goodreads.com/book/show/33573932-the-telescope-i...
What they do it totally out of my field, but I love hearing about it. It's a big source of pride in the university.
Unfortunately, the current hype around neural networks incentivizes scientists to use them even in situations where there is no need. (maybe in this case there's no other way, I'm not sure...)
As you say, it's easy for a neural network to pick up on simulation artifacts, rather than real physical features. The appendix of the original paper [1] explains how they quantify the failure modes of this approach. One kind of cool approach was training an adversarial network to corrupt simulated backgrounds to make them signal-like. The details are sparse, but it sounds like fooling the classifier required a level of corruption that would have been noticeable when comparing the simulation to data.
Can you do this wrong? Sure, but you can also go wrong with the feature engineering that particle physicists have used in cases where they don't trust machine learning. You won't really know if the methodology is sound unless you:
- read the paper
- follow some references (the paper is a short-form "Physical Review Letter", which means it's impossible to give any meaningful review using the text itself)
- hopefully find some more detailed description of what they did
The last point is often the difficult for particle physics collaborations, they are generally a bit protective of their data and internal working.
[1]: https://arxiv.org/abs/2403.02516
The systems folks have had a decade to iron out most of the bugs and there are fairly tightly scoped teams that work on different components. The failure rate of the DOMs is extremely low for something that's deep in the ice and impossible to physically access. They were very well built and designed, but it helps having people on call 24/7.
All the telescopes here have some sort of data center. Icecube happens to have a lot more computers because of the sheer number of detectors, and a desire to filter data so that preliminary/candidate detections can be sent North as soon as possible. Though unlike the other installations, IceCube is far more sensitive to downtime.
The Event Horizon Telescope setup at SPT probably takes the crown for superlatives in terms of data rate, clock accuracy, etc.
It strains credulity to the absolute maximum -- and yet these results are repeated, again and again, all over the world by different groups, using different equipment, based on different detection principles, using different methods.
No idea if this is true but…
Sounds innocent, but it creates a focused directed energy wave from a source array much larger than a measy square kilometer.
Every single time this science hits headlines I have at least (5) seconds of trying to understand how O'Shea Jackson suddenly had a scientific addendum to neutrinos and if that is as-cool or cooler than "No Vaseline".
I really do think it's named that because it looks like a cube and it's made in the ice. So, y'know, icecube.
I take it he's a well known? I've never heard of him.
>They were among the earliest and most significant popularizers and controversial figures of the gangsta rap subgenre, and the group is widely considered one of the greatest and most influential groups in the history of hip hop music
>Rolling Stone ranked N.W.A at number 83 on its list of the "100 Greatest Artists of All Time".
https://youtu.be/5HDgULSW1T8?si=jY6mSLB6jwHMEFT0