rsfern

It is, but NSA reports to the director of national intelligence, not the defense secretary, so it’s unclear (to me at least) that SecDef’s opinion of Anthropic counts for anything here

I guess DOD is large enough they have multiple parallel cabinet level positions

https://en.wikipedia.org/wiki/National_Security_Agency

Admittedly my understanding of QM is a bit vibey but I’ll try to answer

In an atom, angular wavefunctions with wavelengths non-integer divisions of 2pi can’t exist because of the boundary conditions on the wave equation. A free electron can have any wavelength, but once you put it in a box (confine it to the potential around a proton in a Hydrogen atom) the non-integer wavelengths aren’t allowed

I think it’s instructive to think about what the wavefunction represents. It’s square is the electron probability density (technically the wavefunction is complex valued so it’s the wavefunction times it’s complex conjugate). If you have a non-integer multiple wavelength then the wavefunction goes out of phase with its complex conjugate after one period, and if you integrate over the angular domain the electron probability has to be zero everywhere.

This also answers your second question. The radial solution to the wave equation for hydrogen gives you the Laguerre polynomials. They don’t all go to zero at the nucleus though, actually the first one has a maximum at zero because it scales like exp(-r) (See fig 4.10.2 on chem.libretexts linked below). But when you do a volume integral to calculate the electron probability, the probability near the nucleus is low because the integration volume is small even though the wavefunction is large

https://en.wikipedia.org/wiki/Laguerre_polynomials

https://chem.libretexts.org/Courses/University_of_California...

The reason is that electrons (like all quantum mechanical objects) are wavelike. In an isolated hydrogen atom, the electron is in a spherically symmetric environment, so the solutions to the wave equation have to be spherical standing waves, which are the spherical harmonics. The wave frequencies have to be integer divisions of 2pi or else they would destructively interfere. (Technically each solution is a product of a spherical harmonic function and a radial function that describes how fast the electron wave decays vs distance from the nucleus)

What’s interesting is if the environment is not spherically symmetric (consider an electron in a molecule) the solutions to the wave equation (the electronic wave functions) are no longer spherical harmonics, even though we like to approximate them with combinations of spherical harmonic basis functions centered on each nucleus. It’s kind of like standing waves on a circular drum head (hydrogen atom) vs standing waves on an irregular shaped drum head

Of course the nucleus also has a wave nature and in reality this interacts with the electrons, but in chemistry and materials we mostly ignore this and approximate the nucleus like a static point charge from the elctrons perspective because the electrons are so much lighter and faster

While the treatment for methanol poisoning indeed includes ethanol, I don’t think your dosage suggestion is right. Your body would still have to process all the methanol, the job of the ethanol is just to slow down the reaction. If you suspect methanol poisoning you need the hospital, they will administer the ethanol intravenously and I think do dialysis to remove the methanol and the formic acid it metabolizes to (this is one of the toxins in ant venom)

https://doi.org/10.1053/j.ajkd.2016.02.058

There are groups that are actively working on automating conventional labs like this. Most of the efforts I know about use non-humanoid mobile robots or even just a six-axis arm on a rail and some lab space reconfiguration

This issue of accessibility is widely acknowledged in the academic literature, but it doesn’t mean that only large companies are doing good research.

Personally I think this resource mismatch can help drive creative choice of research problems that don’t require massive resources. To misquote Feynman, there’s plenty of room at the bottom

I like this analogy of always choosing “I’m feeling lucky” on Google, I feel like it clarifies a boundary between information retrieval and evaluation that gets blurred by language model summarizations. I’ve been frustrated with the LLM summary at the top of the Google search results for scientific topics because often the sources linked to don’t actually contain the information the summary is citing them for. Then I have a side quest of finding the right backing literature or deciding the summary was just wrong in the first place

I don’t know why the author of the article wrote “could”, but I personally work closely with some non-high-risk-country NIST foreign guest researchers. It’s been filtered down verbally through the management chain that the end of this September is the re-review deadline, and it’s not been stated as a hypothetical.

There is https://orcid.org which is a persistent identifier for a researcher. It would be interesting if sending email to a researchers orcid handle resolved to their current institutional email address I guess?

My usual workflow is find the person on google scholar, find their uni/lab homepage, and hope they published their email there.

It’s all foreign guest researchers by the end of September, high risk countries by the end of March. Your first quote doesn’t imply the NIST sources for this article don’t have firsthand knowledge that this is coming, it’s just that it appears the lab management is avoiding putting things in writing

Not a lot, but what is your point exactly? There are a lot of really Chinese scientists working in the US, and the ones who are postdocs and research scientists at NIST are apparently being pushed out at the end of this month. They’ve already been vetted for security concerns, so that justification is kind of thin.

How many Taiwanese, German, Indian, French, South Korean, etc scientists are working in the US? The ones working at NIST are facing being pushed out at the end of September.

This list of high risk countries is not new (with the exception of maybe Venezuela being recently added, I’m not sure). Researchers with these citizenships have faced extra security review before joining NIST for years, and last year the lab increased the level of security review for everyone (not just this list)

I can understand a clearly communicated need for additional security requirements. But NIST operates almost totally in open science mode, with the main exceptions of being industry cooperative agreements. I don’t think this move to shed international researchers by reneging on commitments from the lab has been at all justified from a security standpoint.

Your first reply was insightful, but this one is not a thoughtful take.

Power consumption and emissions are already increasing, and any regulatory changes in 2025 are not factored in to discussion of those numbers. It’s more interesting to discuss what these changes mean when they are a factor in 2026 and on.

I think this is not quite the right analogy. A better analogy is procedurally generated music, because that’s what model-generated music is. But just like with LLM code generation, the input to the program is natural language (or maybe multimodal image/audio/whatever), and the program is implicitly defined by learning from examples of music.

I think a lot of the issues are the same. Like you might expect the model to go off the rails if you venture away from the bulk of the training distribution. Or maybe the b most effective way to use it creatively is in some kind of interactive workflow revising specific chunks of the project instead of vibe-coding/composing from whole cloth.

I think it’s just scale to the moon rhetoric, like “what if we used 100x more compute?”. Since the units are power and not energy, I’m going with 10 GW continuous load (for training? inference?) but I think it’s not exactly meant literally

I think this is just loose terminology, instead of squaring they should have said “multiply by the complex conjugate”, which is what you do to quantum mechanical wavefunctions to obtain real-valued probability amplitudes

This paper “were RNNs all we needed?” explores this hypothesis a bit, finding that some pre-transformer sequence models can match transformers when trained at appropriate scale. Though they did have to make some modifications to unlock more parallelism

https://arxiv.org/abs/2410.01201

Are there particular libraries that make your setup difficult? I just manually set the index and source following the docs (didn’t know about the auto backend feature) and pin a specific version if I really have to with `uv add “torch==2.4”`. This works pretty well for me for projects that use dgl, which heavily uses C++ extensions and can be pretty finicky about working with particular versions

This is in a conventional HPC environment, and I’ve found it way better than conda since the dependency solves are so much faster and I no longer experience PyTorch silently getting downgraded to cpu version of I install a new library. Maybe I’ve been using conda poorly though?

I don’t think so? the double descent phenomenon also occurs in linear models under the right conditions. My understanding of this is that when the effective model capacity is exactly equal to the information in the dataset, there is only one solution that interpolates the training data perfectly, but when the capacity increases far beyond this there are many such interpolating solutions. Apply enough regularization and you are likely to find an interpolating solution that generalizes well

This is already what the funding agencies do! The merit review process solicits outside expert assessment of the importance, feasibility, and potential impact (including economic development and societal impact) of the research, and the funding agencies do their best to maintain a balanced portfolio of research that is promising for advancing national priorities

By all means we should discuss the transparency of this process, what those national priorities are, and exactly what we (collectively as taxpayers) the risk-reward tradeoff should be. But let’s not pretend that the funding agencies don’t already view science as a public investment, or be too hasty about dismissing the potential medium term economic value of research into for example geology and geochemistry on mars

This is all awesome, but a bit off topic for the thread which focuses on AI for science

The disconnect here is that the cost of iteration is low and it’s relatively easy to verify the quality of a generated C program (does the compiler issue warnings or errors? Does it pass a test suite?) or a recipe (basic experience is probably enough to tell if an ingredient sends out of place or proportions are wildly off)

In science, verifying a prediction is often super difficult and/or expensive because at prediction time we’re trying to shortcut around an expensive or intractable measurement or simulation. Unreliable models can really change the tradeoff point of whether AI accelerates science or just massively inflated the burn rate

“Kill the [model] for trying” kind of sounds like using reinforcement learning to get models to behave a certain way

The article in Soft Matter is open access: https://doi.org/10.1039/D4SM01490A

They have some interesting analysis of the elastic deformation that happens during the rolling process (as opposed to the ball just falling or sliding). Turns out it’s pretty sensitive to the elastic constant of both the ball and the wall

In the case that it’s a human posting sarcastically, wouldn’t that be a false positive?

This is sort of a misguided take IMO - as if LiDAR and other sensor streams are somehow not large scale data while video is

The bitter lesson isn’t fundamentally even about data. the key ingredient is computation (which does scale with data in modern deep learning). There’s even a whole theme on search outperforming learning - until learning methods changed to leverage computation!

I think in a lot of applications, richer data yields better scaling performance. So the bet for self driving is that the complexity of multiple sensor fusion gives a much better constant factor or exponent in the power law scaling of performance with data and compute

Same - I’ve been using einops and jaxtyping together pretty extensively recently and it helps a lot for reading/writing multidimensional array code. Also array_api_compat, the API coverage isn’t perfect but it’s pretty satisfying to write code that works for both PyTorch and numpy arrays

https://docs.kidger.site/jaxtyping/

https://data-apis.org/array-api-compat/

I’ve served as a reviewer for a couple of NSF panels, and one of the things I really liked about the program I reviewed for is that a lot of the proposals included collaborations with local trade and vocational schools to involve and train future technicians and operators in addition to researchers and scientists. I think that’s really important for actually succeeding at the technology transfer goals of NSF, and if I’m reading your comment it does at least partly address delivering direct value for a broader chunk of the population

In the case of the two year old who was removed with their mother, they have US citizen family that the father is trying to transfer custody (it seems he doesn’t have legal status, but I haven’t seen a definitive source)

PDF: https://storage.courtlistener.com/recap/gov.uscourts.lawd.21...

That’s a strawman argument that I would never advocate, and completely ignores my question.

Alternatives include arranging legal custody for the child and to stay in the US with a relative (as one family was attempting), or finding a legal way for them to leave the country with their parents.

Instead, it seems the government is rushing to illegally remove these children before the courts can intervene

One thing I don’t understand is how this is even a choice the parents have the legal right to make, assuming their US citizen children do not have passports (I don’t know if the answer to that is publicly known). Can a child legally be taken out of the country without a passport and some kind of verifications?