omar_a1

I'm not saying the alternative is do nothing. But playing doctor because you think that you're reasonably qualified to administer a ventilator, then we're going to end up with, say 30% mortality rates from patients whose lungs were sucked through a ventilator tube because you guessed the wrong pressure. Or, more realistically, a terrible infection because you decided soap and water in the bathroom sink would be sufficient to clean the apparatus.

And then what do you suppose will happen after 30% of these amateur medical procedures go South? Are you going to throw in the towel, try something even more reckless, or decide that you need to get an actual doctor involved to clean up your mess? At which point, you've just added yet another case to the already overburdened medical system. And at a 30% failure rate, that would become a major burden.

There's a lot of room for action between doing nothing and acting foolishly (such as claiming a 30% mortality rate for botched medical procedures is a rational tradeoff). There are ways to help the situation here that don't involve magically becoming a nurse overnight.

Even using a syringe or blood draw on a patient is something one needs certification to do properly. That isn't bureaucratic nonsense. Having worked with syringes and sterile technique, there are so many ways you can give your patient an infection or otherwise cause them life-altering complications if you don't know what you're doing, even for the actually trivial techniques. Operating a ventilator incorrectly will result in death. Likely a grisly one, as incorrect pressure differentials and human lungs are not a good combination.

Now that doesn't mean we need to stick to these bureaucratic rules as they are for this situation, but the people on here advocating for trying complicated medical procedures while knowing fuck all about medicine (or even biology) really need to take a step back and take an inventory of what they don't know, and can realistically expect to accomplish with neither the technical knowledge nor manual dexterity of a trained clinician. A YouTube video and a technical manual aren't going to cut it, folks.

I'd say to the extent that this behavior contributes to the reproducibility crisis, it has more to do with not backing that exploration with rigorous experimentation.

But more to the point, I'd argue that the lack of funding and publish-or-perish atmosphere has more to do with it. Academics don't necessarily publish because the results are noteworthy, but because they have to, regardless of the robustness of their methodology.

No! It's decidedly not better! This isn't web-scraping where any little bit of information you extract is good, even with a low signal-to-noise ratio. These are human beings. Being wrong here has serious consequences, to the point that you need to weigh the usefulness of the test against the risk of harm of misdiagnosis. You can't just go around diagnosing people willy-nilly if you're not sure about the results!

A false negative means someone will go out with a false sense of security, potentially interacting with vulnerable community members thinking they'll be fine. They may develop further complications due to writing off their symptoms as the flu, and not seek out medical intervention if the situation goes South. A false positive means someone is going into quarantine for two weeks at a major economic and housing consequences (particularly if you want to help poor communities). False positives and false negatives mean more lives will be lost. The stakes are higher, and so too is the need for ethical rigor.

So if we narrow the scope to just testing for Covid-19, are you saying it still takes several years to get to the point of just administering and evaluating the test, right?

I'm not saying the test will take several years to develop, I'm saying it would take you, a self-described layperson, several years to gain the technical skill and know-how to know how to correctly research, design and validate the test.

Specifically, labs are carefully controlled environments. Just getting into the grunt work mechanics of the technique, at bare minimum you need to learn aseptic technique, pipetting, sterilization, reagent prep, handling hazardous chemicals, biohazardous waste disposal, and establishing appropriate environmental controls for your lab. Starting from nothing, those will take at least a year of full-time experience to develop for any semblance of competence. Otherwise you'll be a bull in a china shop; your results will be too tainted by bad technique and inadequate environmental controls to get anywhere.

So is this one of the few areas of human endeavor that is absolutely non-automatable, or can we break down what is/is not within the realm of automation, and possibly leverage the few staff we have relative to the current need?

It's perfectly automatable, but that doesn't mean it's trivial. For one, someone needs to... actually automate it. Someone else needs to implement, maintain and troubleshoot the robotic lab equipment doing the automation. Still require highly specialized skills to do so.

It's kind of like deciding you want to build your own 3 story house by hand. You can do it, but buying a bunch of lumber and tools is a far cry from having the necessary skill to actually build it. Science is hard.

There's also a major ethical issue about the medical implications of testing someone for a stigmatized disease if you're not clear on the validity of the test. False positives and false negatives have a major impact on your community. And hand-waving away the implications of reporting a false positive or negative with the statistical caveats of the results doesn't cut it.

Patients tend to take test results at face value, so the responsibility to of weighing the ethical ramifications of when it's strictly necessary and when it's irresponsible to administer an unreliable test for a given patient fall squarely on your shoulders. (Not that that excuses the current administration's policy of negatively biasing reporting by not testing anyone.)

That's asking for too much upfront before the candidate can even decide if the job is a good fit. Job descriptions alone don't tell you much. That's where the phone interview comes in. It's a two-way street.

Also, from experience, I've never come across a two hour takehome that actually takes two hours. Everyone always says not to spend more than two hours on it, but I can't exactly turn in an incomplete assignment.

I'm glad it is working for you, but it's not something to be proud of. Essentially what you've done is replace the standard 30 minute intro phone interview with a 2 hour SAT.

I didn't think of the solution you had in mind, because I've identified another problem that you didn't consider. And somehow that takes me out of the running. Good luck with that.

Not to mention the nebulous and vaguely discriminatory "cultural fit" problem. So, even if I manage to ace the technical, I still fail.

Yea, if I had also been working on this particular code for the past two years I'd be able to breeze through these problems too.

In other words, incessant, aggressive, and unsolicited, with the reviewer not backing down even after you've implemented their changes? (e.g. Their previously implimented suggestion now are the wrong approach according to their next round of comments.)

I've tried simply going along with it, recognizing my own technical inexperience compared to theirs, but end up getting nothing done because of the capriciously-set goalposts.

This has been a recurring problem for me. Any tips?

...I don't like spiders.

Maybe multiple lenses?

Also, lots of suggestions of incredibly dangerous classroom experiments to jazz things up, coming from people who don't know the dangers of the reactions they're proposing.

This is a bad idea. If you don't do your due diligence, you end up setting school children on fire [0].

[0]http://chemjobber.blogspot.com/2019/06/civil-trial-of-nyc-te...

Ignoring the obvious safety issue of every student having an effervescent, volatile strong acid solution (which is tricky to handle safely even as an experienced chemist), there are a handful of incompatible chemical combinations with aqua regia that would require an immediate lab evacuation and call to the fire department. You'd need constraints and some serious vetting of the students' plans before you could let them touch the stuff.

I like the idea of open-ended experiments, but technique is so crucial to these things that even if they're right in theory, in practice they'd often get a negative result because of technical shortcomings of their technique. This is why you end up with cookbook experiments

It's the difference between writing pseudocode that vaguely looks like it does what you want, and having an actual, executable program that works as intended. The devil's in the details.

You seem overly defensive about your hiring process, and unsatisfied with any answer that asserts men and women are equal. Ironic, no?

> We should be hiring on ability to do the job and nothing else.

This is exactly my point! Yet there is quite a lot of inadvertent, or even blatant, racism and sexism that happens during the hiring process and on the job.

Realistically, the most that hiring managers (save for huge FAANG institutions) can do is thoroughly ensure that their team isn't inadvertently (or blatantly) racist/sexist in their hiring process and on the job, and to post the job in enough places that a diverse applicant pool will see the posting.

With that said, hand-waving away that there are few to no women or African Americans/Latinos/Native Americans/etc. on the team with an overzealous application of the Equality Paradox is a pretty dangerous mindset to get into. It's essentially passing the buck, and is eerily reminiscent of the claims made by 1950's Southern US Politicians that Blacks were the ones who were self-segregating because they wanted to, not the other way around.

What I'm saying is, the ideal 50/50 gender ratio/representative race may be unrealistic for a myriad of reasons, but if you're a 50-person start-up with 2 women, one of whom is HR, and no black people, I'd take a good, hard look at the company culture that's being fostered, and particularly whether turnover for women and People of Color at your company is worse than average.

In other words, ideally the racial and gender distribution of a team would be as inconsequential and unbiased as blood type or handedness, in that the aggregate demographic ratios on your teams would at least match that of the residential population in your area, and ideally that of your broader geographic location.

I'm not doing a good job explaining this clearly, but the simple answer is: more than one. No one wants to be the token hire.

And not to say this happened in your case, but even with that considerable effort, it's still very easy to end up with blind spots in your product that a more diverse team would have caught.

It's the same as hiring for any other level of experience for more routine technical skills. If your team has no experience in this area, they'd need to expend a much greater degree of effort to answer questions that someone who is experienced would already have known the answer to.

But more to your point, if these oh-so-competent ML practitioners were doing their jobs right, we wouldn't be having this discussion.

The whole reason diversity is championed in hiring is precisely because a single individual's perspective can only see so far. And if you have a monoculture team who has experienced very similar life circumstances, you end up with the kind of narrow perspective that leads to more racist soap dispensers.

Basically, your team is composed of white dudes who don't see the problem with a ML training set consisting largely of pictures of white dudes.

To prevent this they'd have needed to A) Employ a black person, and B) Listen to said employee's feedback, in order to recognize the problem.

Edit: Also worth pointing out, just using a representative population sampling would still show racial bias, essentially weighting accuracy with respect to population percent. You'd probably need to have equal samplings of pictures of people from all races/genders/disabilities if you wanted equal accuracy across the board. That also includes picture quality and range of picture quality. Doubling up images, or using corporate headshot white dudes and grainy selfie People of Color could still cause issues.

Same logic applies to labelling. That minimum wage contracting firm used to decide who's who in the photos may exhibit racial bias, by virtue of the fact that most people do. If their accuracy in labelling is racially biased then so too will the algorithms that it's based on.

In short: Racist garbage in, racist garbage out.