CERN and Mistral sign a statement of intent to establish a strategic partnership
home.cern
home.cern
One scenario, and I think what people were paranoid about, is that you could make your cuts really arbitrary, and select a region where you by accident have too few simulation events and too many data events - say "Select all events where the first muon has momentum > 15.6 GeV, but not between 20.1 GeV and 20.3 GeV, unless there are 3 jets, in which case the selection is...". You just try out so many regions until you find something. For this reason you need to keep every step of your analysis justifyable, and a BDT is somewhat more explainable than a neural net (it's essentially just a trained list of simple cuts). Another thing that is frequently done is a blind analysis, where you only look at the real data when everything is finalized and locked in.
There are other issues beyond cherry picking that determine which methods you can use. E.g. the simulation doesn't reflect all correlations between variables, and it doesn't simulate the hardware triggers perfectly - so your simulations might behave different wrt. certain cuts than your data.
Nevertheless, it is not like you are banned-banned from trying different methods. When I was in particle physics (ATLAS and D0), we tried all kinds of creative things including NNs. But you have to convince hundreds of collaborators to put their names under your paper and you have to pass internal peer-review, which can feel a bit conservative and frustrating.
Why did CERN require so much selectivity? Was it a lack of funding? Are particle collisions particularly expensive? Are you competing for "collider time"? I am curious what the limiting factor was.
I think I would lose my mind if I needed hundreds of collaborators to support an experiment.
It was perfectly fine to work on any crazy ideas you wanted to. But if you wanted to publish as part of the CERN collaboration (and if you use CERN non-public data and CERN infrastructure, I'm not sure it was possible to publish as an "independent" anyway), there is a lot of protection to ensure that your article is at a good scientific standard.
As explained in the previous comment, scientists were nervous about "black box spurious pattern", "experimental artifacts" or "elsewhere effect". This was a good thing. The limitations were not there because of budget or access to infrastructure, but because people were conscientious of doing good science.
Remember the NN examples where the NN is given pictures of horses and pictures without horses and had to classify them, and it turned out one criteria was to detect the signature, in the bottom right, of the photographer that specializes in horse pictures? It meant that if you were giving a new picture of a bird but with that signature, the NN would say "it is a horse". Now, how do you make sure that something similar does not occur when treating billions of collision events? The categorized data, especially for the signal, is coming from simulation. What if your NN detects that in the signal, the pixel number 231 is not "hot" like the one in real data (because it turns out that, as sensor can be sometimes, this pixel is hotter in the real detector than it is in the simulation), and uses that to select the signal? Or that the NN overfits part of the data (the majority is classified on sensible cuts, but for more extreme cases, it classifies as signal events where there is an electron with pT between 24.5678 and 24.5685, but as background if it is before or after, because it happens to be what the training data looks like)?
Because of that, there were a lot of cross-checks and precautions being taken. One common approach was to compare more black-box-y methods with other methods that are more easily explainable. That was certainly a good thing. And this was not being seen as a chore: no one was approaching it as if people being good scientists where being personally against your new method. Proving that your new method was checking all the boxes of the scientific qualities was not seen as a bad thing, it was seen as the fun part of building new robust scientific advances.
https://arxiv.org/abs/1207.7214 - The original ATLAS Higgs discovery paper uses neural nets to identify photons. It also uses a neural net b-tagger to improve the H->WW signal to noise ratio.
https://arxiv.org/abs/1303.0763 - used neural nets extra tt+jets noise https://arxiv.org/abs/1503.05066 - used neural nets for signal/background separation.
Um, I believe I have linked you to "actual analysis" where techniques used are more exotic than boosted decision trees.
Join us, in the pit of despair!
It’s one of the coolest environments one can be in. You can righteously wall yourself into a room and solve a problem.
I do hope that this is the start of something cool.
There was a time when CERN would have been establishing partnerships with US labs. The US is not a CERN member but has provided a lot of funding in the past despite it's observer status. These days funding is getting cut, the US is not as trusted as a partner anymore and a European partner is seen as important. This is sad.
That said I'm happy for Mistral and I'm sure they'll be a great partner for CERN.
Makes perfect sense and I’m glad that CERN is focusing on EU providers
It makes more sense if they're the only company who was willing to work on a more custom solution for the proposed LHC model but I don't that a bit as it would be great PR for the frontier labs.
This seems a lot more like an effort to bolster European AI by moving CERN money into Mistral.
What I think the parent post is missing is that breakthroughs in science don't immediately translate into breakthroughs in engineering and even less so into economically viable products.
On the other hand, you could downplay the math findings a bit, since it didn't further our understanding of mathematics just yet, only provided proves to things we didn't have proves for.
Advances in math will probably result in advances in science, which in turn will result in advances in engineering (which is probably where the average person sees an actual benefit in their day to day life). But it's not automatic and a lot of the hand wringing by mathematicians about the AI math results include "is this (AI) method of finding proofs actually going to result in as many of those downstream advances as the old fashioned way?"
(I include theoretical CS as a subfield of math - and believe the S in it is a misnomer).
(in)Conveniently those are also the fields where most ordinary people would experience the most effects from accelerated progress. Dario famously promises potential cures for cancer and dementia.
I don’t think it’s going away, LLMs and other ML stuff will no doubt remain part of people’s workflow, but I also see no evidence for superabundance or anything close to it.
Gerard t' Hoofd (under Martinus Veldman) proved in 1971 that Yang-Mills gauge theories are renormalizable, using a rule-based AI (LISP-styled but manually written in assembly) called "Schoonschip" (I'm guessing called that because there were several failed attempts that went before and failed). This program was an AI in the sense of a search algorithm searching symbolic proofs. It got the 1999 Nobel prize in Physics.
This was hugely important because first, it showed that it was possible to make the Standard Model of physics and it also helped beyond that by showing for a lot of (at that point considered) impossible theories exactly how to make them work (meaning not just a proof that it could be done, but a long list of "if you have problem X, do Y" recipes). One of these became first the "electroweak unification", then the standard model.
Hao Wang proved that the logic of "Principia Mathematica" was actually correct using another symbolic proof generator. His efforts were also the start of many other such programs that contributed lots of stuff to mathematics that nobody sane was ever going to do, such as finding all the higher-order versions of the quadratic formula (there's a lot of them).
At Stanford university, tens of thousands of molecules were found in 1965 by having an AI, this one was what's called an "expert system", evaluate mass spectroscopy data and combine it with chemistry knowledge in the form of rules about what was possible or not possible and if only one option remained, spit it out. It almost doubled the sum total of our chemistry knowledge if counted by known molecular structures. It was later adapted to make mass spectroscopy useful in finding the composition of samples.
Then there's Alphafold, first a ConvNet (Resnet+Convnet to be exact), then switched to the transformer architecture, which gave use the folding of just about every protein we have ever found, and even more proteins that we've never even found. Nobel prize in Chemistry 2024.
Less modern AI (aka. machine learning) is used all the time in science. I don't know of any commercial breakthrough yet, but there are some promising developments, like in antibiotics. If it materializes, that would be a very big deal.
As opposed to enshitifying the lives of the working class in most other areas.
Today they’re one of the biggest compute in the world, they have the strongest data interconnects, multi gigawatts of power and their deeply embedded in science thousands of the greatest scientist in the world work there including thousands on machine learning.
In my opinion, high energy particle physics is a dead end yet still they raise money like nobody in Europe for the next future circular collider they’re asking for close to €100 billion
Clearly, there is a path there where a fraction of that money, infra, and a fraction of the incredible geniouses that work there could be rerouted into a Opus/Mythos/Astra shape open / European system
Why that? I'm not following all that too closely
What? It’s not that kind of nuclear. The members collaborate because it’s insanely expensive to build these huge accelerators and not many countries have internal pools of scientists big enough to justify having one.
> Today they’re one of the biggest compute in the world
Not really. They have cool computers but the biggest are in the US and China.
> In my opinion, high energy particle physics is a dead end
Out of curiosity, what do you base this opinion on?
> Clearly, there is a path there where a fraction of that money, infra, and a fraction of the incredible geniouses that work there could be rerouted into a Opus/Mythos/Astra shape open / European system
No. People, even smart ones, are not interchangeable, universal geniuses. They end up in one position because they are good at something, it does not mean that they would be productive in a different position. And understanding our universe better is well worth the pennies we spend on CERN. Other people are already working on LLMs.
Besides, it’s not at all what is happening here.