I’m in the building sciences. The biggest unanswered question we come up against almost daily is “what the fuck was the last guy thinking?”. And we avoid, daily, admitting we were the last guy somewhere else.
This sounds like software engineering in a nutshell.
Software engineering is the study of constantly calling your predecessor an idiot.
Rules of Tech Support:
Rule T18 - You are incompetent. You just don’t know it. At least, that’s what your replacement will think.
Rule T18A - You will have to deal with techs who are incompetent.
Rule T18B - Sometimes, you really are incompetent.
Especially when you are that predecessor
I feel inappropriate near all the very universal questions here, but as a paleontologist specialised in some reptilian groups, the question would probably be “where the fuck do turtles come from?!” The thing is that fossil evidence points to different answers when compared to genetic evidence, and thez separated long enough from other extant groups that we keep on having new “definitive” answers every year
Genomics makes this answerable though? It’s just a matter of whether DNA is preserved or not in fossils. Genomics is more reliable than comparive anatomy. Comparative genomics can accurately place turtles in animal phylogeny. Sorry if I misunderstood your post. Or am I wrong here?
In phylogeny, genomic is just another tool. The point is that turtles are os course animals, but they do branch off of different reptile groups if you look at morphological evidence (which includes fossil data) or at molecular (genetic) evidence (which only includes extant species). This is not something frequent, as usually molecular evidence tends to strengthen previous morphologically established evolutionary relationships. And even though molecularists are more numerous today, their methods are neither better or worse than anatomy.
Phylogeny is not as straightforward as some people make it seem, and especially molecular phylogeny tends to rely on abstract concepts that can’t always be backed up by biological evidence (I’m not saying it’s wrong, it’s very often very good, juste that a lot of people doing it do not understand the way it works, and thus can’t examine the process critically).
And so turtles’ origin are still very much an active debate!
Maybe we’re not talking about the same thing? I was thinking about the diapsid debate, where genetic evidence is overwhelmingly strong in favour for diapsid evolution Mitochondrial DNA evidence: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC24355/ Micro RNA evidence: https://pubmed.ncbi.nlm.nih.gov/21775315/
Tbh a core multi gene ML tree to all other reptiles would prove it beyond a shadow of a doubt, maybe someone’s done that already but I haven’t been able to find it.
The diapsid part is very likely indeed, as fossil skulls of early stem turtles do show some temporal openings ( https://www.annualreviews.org/content/journals/10.1146/annurev-ecolsys-110218-024746 ) The point is more where do they nest within Diapsida, more closely to the Lepidosauromorpha, or to the Archosauromorpha, and where precisely if within one of those clades. The point is that can’t quite be proven using only extant species, whether by DNA or morphological evidence. And concerning ML, the methodology is often criticised, not because it’s bad, but because it’s opaque and thus it is difficult to justify and understand as a process
How to get supervisors, superintendents, school boards, and even politicians to let teachers teach. It’s understood that overtesting reduces learning. It’s understood that rigid curriculums don’t work, and you really should be tailoring lessons to the capabilities of the class. All kinds of educational philosophy is understood well and in depth… but being permitted to apply any of it?
As someone who does hiring for tech, the problem is things are metric driven. You can’t extract metrics from letting teachers “teach their own way” without standardized tests, and if you don’t have metrics, you don’t know if “teaching their own way” is working in practice (you can extend this logic down to understand the rigid ciriculums).
By the way, I think this is all bullshit, but that’s why
Oh yeah, I fully understand why the stupidity happens/happened. I don’t know how to fix it or if it can be fixed… that’s why I posted it here, in the unsolved problems in your field thread!
I have a question about rigid curriculums. This is mostly for high school. Many of my teachers had curriculums and syllabi that they had been using for years and kept them basically the same, and then there were the AP classes where the curriculum was determined by the AP exam. I felt that I learned really well in AP classes and we would get through much more advanced material in the AP classes than in others. And I also felt that the teachers who had developed somewhat fixed curriculums from experience taught much more efficiently than those who hadn’t. It never felt like the teachers were changing their curriculum for each class whether it was an AP class or not because most had their curriculums kind of figured out over the course of teaching for many years. And most of the teachers I had in high school were excellent. So my question is, why is it believed that rigid curriculums don’t work? Because in my schooling experience, whether the rigid curriculum was developed by the individual teacher or by an external organization (like AP), the class seemed to benefit from having fixed goals for the year.
Is P, NP?
Guys I swear this actually makes sense…
I solved this in undergrad.
P = NP when N=1. I don’t understand what the big deal is.
I don’t think you passed
Also P is 0 when N is 0
Trying to prevent bacteria from developing antimicrobial resistance. At these rates in 30 years antimicrobial resistant bacteria are projected to kill more people than cancer.
Clearly you need to figure out how to give antibiotic resistant bacteria cancer.
Uncontrolled dividing of the most dangerous bacterias known to man? What could go wrong?
I’ve been around the AMR space for a while, but only as a collaborator. Have helped do some bacterial assemblies and help find methods of detecting ICE. I’m a bioinformatician so I get to jump onto a bunch of different projects.
AMR is scary and not really in the public knowledge of upcoming issues. I think about it every time my son had an infection while he was very young and hope he didn’t get a resistant strain.
So are there any good news in this respect?
There was a paper back in December about a new class of antibiotics being discovered thanks to the use of Deep Learning.
This looks like a decent writeup about it, the paper itself is not open access
This is very welcome as it has been a long time since the last new class of antibiotics was discovered. Here’s a good paper that talks about the timeline of antibiotics
It’s been a little while since I took the AMR course, so I’ll let the papers speak for themselves instead of trying to quiz myself here on Lemmy.
How much of this resistance is down to feeding livestock antibiotics compared to doctors over-prescribing to people, or what is the cause do you know? Is there any way to slow down the rate?
The level of AB use in livestock in various countries is astonishing.
Most european nations have to keep a very strict log of which antibiotics are used, and for what reason.
Meanwhile, until recently India was using Colistin as a growth promoter.I saw numbers on this recently. It was something like 80-90% of all antibiotics are given to livestock. So this is a huge contributor.
I think there are so many new and great ideas in this space but you have to consider how science is funded. Funding bodies and reviewers want incremental research that is safe. This has led to our current situation. Phage therapy has been around for so long but is only in the last 10 years gained creditability and treated as a path to take. Ultimately, antimicrobial resistance is incredibly solvable even at a policy level and definitely across many scientific levels. But it requires more cooperation than farms, pharmacies, hospitals, states and countries can muster.
We really need a big push into bacteriophage research I think. Get the bugs all killing each other so we can keep our antibiotics for emergencies.
I’m an IT auditor. “What the fuck?” is the main question, we ask it daily
I do other audits, mostly safety and environmental, and my big question is usually “nobody made you write this, why would you write this down if you don’t want to do it?”
Oh so so much of "dude you mande this rule up, you reviewd this document, why is this process nothing like this?!
Can you explain? Are you referring to catching people doing stuff they shouldn’t have been doing?
For most regulations, the laws and rules say something like “companies must ensure X doesn’t happen”, and the companies themselves have to come up with a way to do that.
Let’s say the law says “companies that transport apples must be able to show which batch went where”.
Company A says “to comply with the law, whenever we move a shipment, we store the shipping order on our computers”
Company B says “to comply with the law, the truckdriver will film the place they left, count the apples when leaving, then email the entire dashcam trip, and count the apples on arrival”.
Neither process is wrong, they both follow the law. But when I go to Company B, I promise you they’re going to fail the audit. They’re (probably) not doing anything illegal, but they’re going to fail their audit because no truckdriver is going to count a truck full of apples.
They made that rule, and they really didn’t have to.
It’d be interesting to see that answered scientifically.
I work in IT and haven’t had to go through an audit yet knocks wood
Any war stories you can share?
Mostly cybersecurity strugles. If you invest millons in a castle with a gigantic lock and a pit full of piranas, would you leave the service entrance open and give everyone in town the key? Yeah, more commom than not.
But an IT audit is only necessary if your company goes public or is the owner wants it, maybe if you are a tech company.
When I was a graduate student, I studied magnetism in massive stars. Lower mass stars (like our sun) demonstrate convection in their outermost layers, which creates turbulent magnetic fields. About 1 in 10 higher mass stars (more than ~8x the mass of the sun) host magnetic fields that are strong and very stable. These stars do not have convection in their outer layers (and thus can’t generate magnetic fields in the same fashion as the sun), and it is thought that these fields are formed very early in the star’s life. Despite much effort, we haven’t really figured out how that happens.
I love how you stopped to explain stuff like what a big star is, but not the funny magnetism itself
My brother works in molecular biology; he tells me the field’s understanding of peptides have only just begun and it’s only through machine learning that they are now starting to make progress. 99% seem to be post-translational garbage, the other 1% is likely to be the basis of a revolution of treatment options.
How does immunology work?
Pro tip: nobody understands immunology and anyone who tells you otherwise is lying
After covid, this strikes me as a dangerous thing to say. Are you an immunologist and could you expound on this?
My field of expertise is bacterial pathogenesis with a particular interest in pneumococcal pneumonia.
And it’s true, immunology is ridiculously complex that no one person can ever hope to fully understand it. Immune cells are helpful or detrimental depending on the context, and sometimes even both. And we don’t really fully know why. The problem is that pathogens and humans have been in an evolutionary arms race for billions of years, and unraveling all of that evolutionary technical debt is Fun™
To give an example, Toll-like receptors are one of the most important pathogen-detection mechanisms, and they were discovered just about 25 years ago and people only really figured out their importance about 20 years ago. There are researchers who have spent the majority of their careers before the discovery of one of the most crucial immune pathways.
We really don’t know what’s going on with immunology and to say otherwise is, as I’ve said, an outright lie. People seem to overestimate how much we know about the immune system, not knowing that we are still very much in the “baby phase” of immune research. The fact that we are able to do so much already is really kind of a testament to human ingenuity than anything
My personal experience is that people who claim to know completely about how the immune system works is more likely to be a science denier (or more likely, naive)
Thanks, that was a great answer! I had no idea it was so complicated. I was definitely in the naive camp there.
Ya’ll think you have real unsolved problems. I’m here with “naming variables” (⌐■_■).
As a software engineering researcher, I strongly agree. SE research has studied code comprehension for more than 40 years, but for that amount of time, we know surprisingly little about what makes really high-quality code. We are decent in saying what makes very bad code, though, but beyond extreme cases, it’s hard to come to fairly general statements.
Genuinely curious - what do we know makes code very bad?
A few bad things in code for which we have fairly consistent evidence:
- high nesting depth
- meaningless or single-letter variable names
- lots of code duplication
- very inconsistent formatting
- very complicated Boolean conditions with AND and OR
- functions with a lot of parameters
we become programmers because we lack creativity. my brain short circuits when i have to come up with something other than “foo”, “bar”, or maybe even “baz”
Programming is quite literally creative problem solving, so I doubt that programmers lack creativity.
Problem solving, of course, but creative writing, composition, and art… not my cup of tea.
I have the opposite problem, my variables are sometimes too descriptive. I even annoy myself at times with VariableThatDoesThisOneThing and VariableThatDoesDifferentThing just because I want to be able to come back later and not wonder what I was smoking.
If the solution to a problem is easy to check for correctness, must the problem be easy to solve?
For instance, it is easy to check if a filled sudoku grid is a valid solution. Must it therefore be easy to solve sudokus?
Most people would probably intuitively answer “no”, and most computer scientists agree, but this has still not been proven, so we actually don’t know.
That’s actually the simplest and clearest description of the P/NP problem I’ve ever read.
Well, there’s counterfactual examples of this, so it must not be true.
In pretty much every single relationship worldwide, one person can very easily determine if the recommendation from the other for where to eat or what to watch is correct or not.
And yet successfully figuring out where to eat or what to watch is nigh impossible.
Isn’t it proof enough? Using the Sudoku example: there are certainly different levels of difficulties, depending on how many numbers are set in the beginning and other parameters. Checking if the solved answer is correct, is always the same “difficulty” - thus there is no correlation between the difficulty of the puzzle at the beginning and checking the Correctness. Some people might not be able to solve it, but they certainly can check if the solution is right
Isn’t it proof enough?
Unfortunately no. The question is a simplification of the P versus NP problem.
The problem lies in having to prove that no method exists that is easy. How do you prove that no matter what method you use to solve the sudoku, it can never be done easily? You’ll need to somehow prove that no such method exists, but that is rather hard. In principle, it could be that there is some undiscovered easy way to solve sudokus that we don’t know about yet.
I’m using sudokus as an example here, but it could be a generic problem. There’s also a certain formalism about what “easy” means but I won’t get into it further, it is a rather complicated area.
Interestingly, it involves formal languages a lot, which is funny as you wouldn’t think computer science and linguistics have a lot in common, but they do in a lot of ways actually.
You can solve any sudoku easily by trying every possible combination and seeing if they are correct. It’ll take a long time, but it’s fairly easy.
Well it just so happens that the definition of “easy” in the actual problem is essentially “fast”. So under that definition, checking every single possible solution is not an “easy” method.
Probably not the most complex, but in programming, the salesman problem: intuitive for humans, really tough for programming. It highlights how sophisticated our brains are with certain tasks, and what we take for granted.
Also, related xkcd.
I once accidentally worked myself into trying to solve the traveling salesman problem. I was doing some work on a very specific problem, and I got to a point where I couldn’t figure out a way to efficiently link up a bunch of points. The funny thing is that I knew about the TSP, but I just didn’t realize that the problem I was trying to solve was a case of the TSP. After a couple of days trying to figure it out, I realized what it was, and that it was futile.
It was a good lesson to always try to find the most abstracted version of the problem you are trying to solve cause someone smarter has either tried and failed or tried and succeeded.
I think it’s dark matter. There are so extremely many theories around it and it’s very hard to measure experimentary.
No one really knows how osmosis works.
I had to derive osmotic pressure for my statistical mechanics exam in my bachelor’s. So in what sense don’t we know?
We don’t know what molecular mechanism creates the pressure.
Isn’t it just water carrying stuff across a concentration gradient?
Gravity is just stuff falling, but there’s complexity there, as well.
No? Water can’t carry anything across the membrane.
Oh. Then how does the stuff get across the membrane? Does the membrane allow stuff to move over a concentration gradient without the water?
How to accurately estimate signal crosstalk and power delivery performance without FEM/MoM simulators.
For people and companies that can’t afford 25k-300k per year in licence and compute costs, there is yet to be a good standard way to estimate EM performance. Not to mention dedicated simulation machines needed.
That’s why these companies can charge so damn much. The systems are so complex that making a ton of assumptions to pump out some things by hand or with bulk circuit simulators often doesn’t even get close to real world performance.
If someone figured out an accurate method without those simulations, the industry could also save a shit ton of compute power and time.
