While I strongly empathize with the author's feelings on this (I've had similar feelings in my own field), I also want to voice an opposing viewpoint:
> Good/useful/valuable/important/positive-ROI science doesn't necessarily require everyone to know what the major discoveries are nor even care when 60% of studies fail to replicate.
Handful of reasons I feel this way:
- If I care deeply about exactly 10 of the 100 publications in my sub-field, and you care about a different 10, it may not matter much to either of us when 60 of them are later refuted. I may have not cared about those specific conclusions, already been skeptical about them, or have several other studies and my own unpublished results to maintain my confidence in the broader idea.
- While we all have great examples of dramatic upheavals in _other_ people's fields — pick your favorite of cosmology, genetic engineering, mathematics, etc. — when you're immersed in it, science is much more incremental, subtle, and complex. Congratulations! You've ventured on beyond the Dunning-Kruger effect. Scientific progress is not a series of miracles.
- Relatedly, I'm guessing most scientists are much more aware of the shortcomings in their own and their peers' research. Do keep this in mind while reading the perspective of an insider. Be skeptical, by all means, but not only of psychology.
I don't know this particular story, but — FWIW — it's common for different teams working in the same field to be aware of each other's manuscripts and impending submissions. When two teams are about to submit papers with the same discovery, they'll often work with an editor to put it in the same journal issue.
By coordinating a simultaneous publication they can get extra publicity for the discovery, both get the first-mover advantage in citations (both papers get cited by everyone), and also get breathing room to be fully rigorous and write the best possible paper.
I too owe a lot of my PhD and postdoc productivity to Snakemake. It's my bioinformatics super-power, allowing me to run a complex analysis, including downloading containers (Singularity/Apptainer) and other dependencies (conda), with one command.
Great for reproducibility. Great for development. Great for scaling analyses.
I think my gripe with the argument in this video (which is, nonetheless, interesting from an offensive marketing perspective) is that equating making clothes out of petroleum products with "being a friend of the oil and gas industry" is disingenuous. For the most part, the environmentally disastrous emissions come from _burning_ fossil fuels; I'm guessing cheap plastics are just a side-effect of a massive energy market, and not a major driver of the industry.
I just tested the PDF reader out for about 5 minutes and I can already say it's substantially better for me on MacOS than any other PDF reader besides Preview. No glitching like Acrobat and the annotations features are more inline with what I want than Preview's are.
(Responding more to the title than the well written article itself:)
The question is never what would be good to add to a curriculum; the question is what we should remove.
Students _could_ have classes on statistics, critical thinking, physical education, some sort of art, the "classics", religion, emotional intelligence, resilience, entrepreneurship, ...
Share a point-of-view on which parts of the high school curriculum should be dropped to make room for these and then we can have a _real_ debate (slash argument).
> > > Drew Weissman working at the _privately_ funded University of Pennsylvania picked her up. You wouldn't have even needed to look up if the University of Pennsylvania was privately funded, because just about every important breakthrough in science in the last 30 years occurred at a privately funded American university.
> >1. Drew Weissman's research was (almost certainly) majority funded [by public money]...
> At least where I live, many government grants are only available to people who have also managed to get private industry funding for their work too.
I don't believe this is true at Penn, or other universities that I'm aware of in the US. I don't doubt that Dr. Weissman is able to get private foundation funding or corporate sponsorship, but that's very likely an effect of his private funding, rather than the other way around.
> > > something like 10 privately funded American universities make more scientific breakthroughs than the 10,000 publicly funded universities around the world put together
> >2. Also very false, although arguably hard to prove one way or the other. [... How do you define breakthroughs? Press releases from university PR departments?]
> You have to understand that it is in the interest of the tens of thousands of people doing work doing non-sense research to pretend their research is important. Just because you hear about them telling you how important their worki is in the media, doesn't mean it is.
Yes, this is basically my point. Just because you can bring to mind "breakthroughs" from major private universities, doesn't mean that most valuable research is conducted there.
> Anyone who has actually worked it research knows every field is filled with 10's of thousands of garbage research papers that are of no value, and that all the key work is produced by just a hand full of people.
As someone who works in research I agree that there are many garbage papers. I absolutely do not agree that the key work is done by a handful of people...unless hundreds or thousands of productive researchers (depending on the field) counts as a handful.
> I remember also reading some researchers that looked at dozens of fields and breakthroughs and found the same thing. All the real work in any breakthrough is done by just 2 or 3 people at most. so this is incorrect, what I said is actually very provable.
I still don't know how you're defining breakthroughs. I suspect that you are referring to how much work any of the (let's say 10) authors on a single paper did in that publication. Yeah, 1-3 people on any given paper sounds about right. But most/all of those (7-9) other folks are the key people on other publication. This does not mean that most researchers are not contributing to science (although how we allocate credit in authorship could certainly use some fixing).
> > > Regardless, this idea would be impossible in the first place at a government funded institution so why even bother mentioning it.
> > I'll point out that private money can go to diverse people at diverse institutions (including government funded universities).
> Yes, that is my point...? You are calling my comment a knee jerk reaction yet you have responded without seeming to understand any of it.
I read this part of your comment (and, in fact, all of it) as saying that public universities are inferior because people working there cannot pursue ideas that the _government_ deems unacceptable because they are government funded. Throughout my response, I'm pointing out that the public/private status of the university has little bearing on where researchers' funding comes from. Not only do most researchers at private universities get public funding, but there's nothing stopping researchers at public universities from receiving private funding.
I believe that you have a fundamentally incorrect understanding of how research and research funding (in the US) works.
> just about every important breakthrough in science in the last 30 years occurred at a privately funded American university.
Really makes it difficult to take the rest of this comment seriously.
In case the ways it is "not even wrong" need to be detailed:
1. Drew Weissman's research was (almost certainly) majority funded by public money doled out by the NIH, NSF, and other national organizations. The public/private status of the university has little bearing on that, as most university research funding comes through these agencies, with something like 50% generally going directly to the universities themselves. Research at "private" universities as it currently exists would not survive without this mechanism.
> something like 10 privately funded American universities make more scientific breakthroughs than the 10,000 publicly funded universities around the world put together
2. Also very false, although arguably hard to prove one way or the other. How do you define breakthroughs? Press releases from university PR departments? Patents? Either way (or by some third—hopefully measurable—way that I'll allow you to define for us) I guarantee that you need to go much further down the list of private universities before you match the output of "all publicly funded universities around the world put together".
I imagine that you have in mind important (and/or well publicized) advancements from MIT/Stanford/Harvard and are forgetting about the enormous amount of research output from public universities (which include but are not limited to Berkeley, CalTech, U of Michigan, Georgia Tech, Virginia Tech, U of Texas, Ohio State, etc.)
> Regardless, this idea would be impossible in the first place at a government funded institution so why even bother mentioning it.
3. As you hardly cite any evidence for this, I'll point out that private money can go to diverse people at diverse institutions (including government funded universities).
So my question to _you_ is whether this comment is motivated by a knee-jerk anti-government reaction, or if I'm entirely misunderstanding where you got these ideas?
I'm not sure if it's related, but I've seen discussions of modern ML methods (in particular those trained using stochastic algorithms...maybe also models with low float precision...?) approximating Bayesian methods. The way I've imagined it is that the training path, by virtue of its stochasticity, resembles MCMC sampling and therefore tends to end up in regions of high posterior volume (the "typical set"), rather than high posterior density. I could see this resulting in a fit with parameters closer to their conditional expectations (in the Bayesian sense), which should be more generalizable to new data, hence fewer issues with overfitting.
A consequence of this would be that if somehow a method were able to successfully find the _global_ loss-function minimum on the training data, it would perform worse on the the test set. Fortunately, our optimization methods _don't_ find the global minimum at all.
Can anybody point me to literature on this idea? I don't know if my uninformed interpretation is actually close to what experts are thinking.
This is a really phenomenal explanation; took me from a vague sense of Bézier curves beings some sort of interpolation between points to actual intuition about what's going on.
I'm not interested in game dev at all, but now I'm curious if Bézier functions show up in topics that are important to me e.g. the numerical integrators used in the No-U-Turn sampler for Bayesian inference...?