ylem

Really cool! What problem are you working on? I live on the experimental side. At least in condensed matter, there are people having fun on the theory side as well.

I agree with you on UQ. For example, I have seen a couple of talks in my field of neutron scattering, where people are using denoising autoencooders to remove artifacts and fit data. It's also clear that no one has any idea how this effects the uncertainties on parameters for models that are fit on the denoised data, much less what happens if the models are not appropriate for the data.

I think reproducibility can be tackled--at least some journals (shameless plug--I'm a lowly associate editor on science advance) are strongly encouraging people include data/code with publications. I have reviewed papers in Nature Comput. Materials where people have included data/jupyter notebooks (not perfect, but a very good start). It would be great if funding agencies started adding more teeth to requirements on data sharing. However, many more groups are putting their code on Github.

There are a lot of cool advances in AI and physics. In my particular field of condensed matter physics, a number come to mind. One is trying to automatically extract synthesis recipes from the literature. Imagine that you want to see how people have synthesized a given solid state compound. Then searching through the literature can be painful. A great collaboration from MIT/Berkeley did this using NLP. I don't know what blood oaths they signed, but they were able to obtain a huge corpus of articles. But, how to know if an article contains a synthesis recipe? They set up their internal version of Mechanical Turk and had their students label a number of articles. Then they had to find the recipes, represent them as a DAG, etc. They have now incorporated the result with the Materials project (https://materialsproject.org/apps/synthesis/#).

There are groups that are using graph neural networks to understand statistical mechanics and microscopy. There are also a number of groups working on trying to automate synthesis (most of it is Gaussian process based, a handful of us are trying reinforcement learning--it's painful). On the theory side, there is work speeding up simulation efforts (ex. DFT functionals) as well as determining if models and experiment agree (Eun Ah Kim rocks!).

Outside of my field, there has been a push with Lagrangian/Hamiltonian NNs that is really cool in that you get interpretability for "free" when you encode physics into the structure of the network. Back to my field, Patrick Riley (Google) has played with this in the context of encoding symmetries in a material into the structure of NNs.

There are of course challenges. In some fields, there is a huge amount of data--in others, we have relatively small data, but rich models. There are questions on what are the correct representations to use. Not to mention the usual issues of trust/interpretability. There's also a question of talent given opportunities in industry.

Shameless plug--The American Physical Society has a topical group on Data Science. Since our annual meeting was cancelled due to Covid, we've been running a free series of webinars on data science and physics: https://www.youtube.com/channel/UCfPG-nSsgnFeWuzgPcbKlCw/vid...

If anyone is interested, we have one on data science in industry coming up: https://attendee.gotowebinar.com/register/604483936035643777...

I agree with you on being interviewed. Bringing someone onto your podcast is more of an implicit endorsement unless you actively push back on their other views. If you don't know about those views, you should have done your research. But, if that was the case, he could have stated it. Imagine that he invited on a vocal flat earther, but discussed gardening with him. In his position, it raises doubts about judgement. If he says, "I didn't realize he was a flat earther--the earth is not flat", he probably keeps his job. But if he threatens to sue people who talk about his appearance on the show, it's a different story.

You mean submit a paper?

Who?

He raises a good point that appearing on a podcast is a red herring. Suppose he actually just discussed say physics on the podcast? It would be more a question of what he said. Or, if he just said, "whoops, I had no idea who this guy was". Who he chooses to host on his own podcast is a different story.

Alexander?

Sorry to hear that. How much power did he actually have to damage someone's research career?

I asked this earlier, but are these out of context questions? I am not familiar with the guy: https://www.splcenter.org/fighting-hate/extremist-files/indi...

Are these comments taken out of context? https://www.splcenter.org/fighting-hate/extremist-files/indi...

He seems out of step with scientific consensus--but he wasn't removed as faculty, but rather from an admin position.

Maybe I missed something. Did anyone threaten his tenure? I would distinguish say an admin appointment from a research appointment.

The materials project has experimental and theoretical structures. On the experimental side, how does it compare to ICSD? How up to date is it in comparison? I love the API for the materials project!

I think that's where commercial programs make their money--in phase identification. JADE comes to mind

There's the classic, Protein Data Bank. A number of people in the field saw it as a moral imperative. It's interesting how open different disciplines are to openness

I was just using it today for a quick calculation!

Isn't this a one year experiment due to Covid?

Why do you think this is a bad idea? Shouldn't it be a data based decision? https://advances.sciencemag.org/content/5/1/eaat7550

Thanks for the excellent response! I have looked at this in the context of physics GREs and it doesn't seem to be terribly predictive--but I think that graduate admissions probably has fewer candidates than undergraduate admissions. So, is there time/resources for more holistic admission at the undergraduate level (at least for the first pass)?

Ok, so you think that say some state schools it might work ok, but not at say Stanford or Harvard. What about say folding in AP exam scores?

These are all excellent points! But, I think this is the discussion we should be having. I doubt we can get the data, but universities have data on both the performance of students they have admitted (along with their schools) as well as their performance at the university. It seems like they could do their own analysis.

You mentioned the question of what the A means. I wonder if just the act of working to get the A tells us something and that something may be more predictive of college success than the SAT/ACT.

Or, another thing to imagine looking at would be say AP exam scores. I am biased, but I would imagine these would be more predictive than SAT/ACT and probably more informative than GPA.

The question of change in major is interesting--but I think even the coarse question of say graduates/doesn't graduate would be interesting to see how it correlates with high school GPA (as compared to SAT/ACT) before delving into the more difficult cases that you rase.

I'm a bit confused, but isn't this a one year response to the effects of Covid? I saw that there were difficulties even with AP exams this year.

This is an excellent question--that should be answered by data. Is there a thresholding effect? It could be that say there's not much difference in a hundred points in the SAT. But, maybe GPAs actually are more predictive. I assume that rec letters are useless to first approximation unless the recommender writes something about the student that raises flags.

It's often good to question assumptions. I happened to do well on SATs/ACTs, so I could be biased in their favor. But, I was curious--what is the actual predictive value of SATs/ACTs on college outcomes such as say GPA or graduation rate. This study would suggest that it's actually more predictive (at least in Alaska) than test scores: https://ies.ed.gov/ncee/edlabs/projects/project.asp?projectI...

Has anyone else found other studies? I would be shocked if there isn't some kind of meta-analysis. Before getting into any discussion about social issues, let's see whether the tests are actually predictive. Or, if it's just a cargo cult.

My intuition was that these tests (ACT/SAT) would help to normalize for different levels of schools, but I will go with data over intuition.

In some sense, this move by Harvard will allow for a natural experiment to see if there are actually changes in outcomes when GPAs are/are not considered.

Dumb question, but are GPAs useless? I mean a plot of say high school GPA vs say college graduation rate or college GPA would be rather informative.

I used Visual Group theory as the basis for some workshops. It's a really good read.

Hi! I am not sure about your background--which matters for explanation. I cannot speak to 6 feet in particular, but the basic idea starts with the assumption that the primary vector is large respiratory droplets. If that's the case, imagine you are a droplet. You are acted upon by gravity and by air/temperature. As you move along, you will fall. At some distance, most droplets above a certain size will have settled. That's the basic reasoning. You can argue about the particular cutoff, but the bigger question is how relevant aerosols are. But, the problem is that in the early spread of a virus, the growth in infections is very fast. Someone has to guess. If you wait for perfect knowledge, a lot of people will die because you are taking a linear approach to an exponentiatial problem.

I think one challenge is that pre-print servers are meant for scientists, not the public. Take physics. If I read a paper on the arxiv in my niche, I can evaluate it's quality. I often referee papers. For me, it's useful to get the early result and if there are gaps, then it's on me to catch them. For those working on Covid19, they are experts, writing for experts. Someone may put out a paper about aerosols in enclosed spaces and how they linger. Someone else might write about how this doesn't verify if the virus RNA detected indicates a live virus. This is how science works and you are seeing it at an accelerated pace. It's actually rather beautiful. But, if you are say a journalist and don't understand this, you can see an article and write a story about it, without understanding that it is preliminary. In short (too late for that), the problem is not that articles on pre-print servers disagree ad we learn more (that's also science)--it is failure of reporting to inform people that these are preliminary results and could be wrong.