1,687 karma · joined May 1, 2012
[ my public key: https://keybase.io/klmr; my proof: https://keybase.io/klmr/sigs/F9zackvFn6PE00C5T_IlAkRQPH_CVvp0aqSA-hysJ30 ]
That said, my profile is very clearly linked to a real person, I don’t see how resale would even work (not that I’d consider it).
Packaging and notarisation is a pain, but it is possible, even though the app we’re distributing is self-contained, so it includes a minimal JRE distribution, dynamic libraries, utility binaries, and a loader. All of these components must be code-signed and notarised (not individually, only the complete bundle is submitted, but all components are inspected).
The interview was for a senior position.
> Even if you do learn it all, people tend to lose knowledge they don't use
Let me emphasise that my interview was not a knowledge test (and at any rate I don’t study for interviews). I wasn’t expected to know by heart how to implement popcount. The interviewer was trying to see me work. Successfully, I might add. — Another question I got concerned something I had no knowledge of, and I had to derive a solution myself. In fact, I failed to do so, but that didn’t prevent me from getting the job since the interviewer was satisfied with what they observed about my thought process.
And as for coughing etiquette, I can’t remember when I last heard somebody cough or sneeze when out. I’m sure it happens (after all, it’s sometimes unavoidable) but it has definitely gotten much rarer. In sum, it’s a very strong claim to say that “masks are a pointless waste of time”, and almost certainly wildly off the mark.
^1 Of course hand-washing should be practiced nevertheless, because it also prevents other infections, notably GI infections. But paradoxically excessive hand-washing also has negative health implications: soap and alcohol destroy the skin’s protective layer and dry it out, and the mechanical action of hand-washing exacerbates this through abrasion, which causes micro-lesions, irritation, and increases the risk of catching infections. And unlike for mask-wearing we actually have very good evidence for this. We don’t yet know the exact consequences, but the NHS has noticed a strong uptick in dermatological conditions since the beginning of the pandemic, some of which are serious. (Here’s a publication on this: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7102542/)
That’s wrong, the opposite is the case. The study controls for this by looking at a synthetic control over the same time period, which consists of other populations without mandatory mask-wearing. It’s an observational study so it can’t control for everything but it does control for seasonality, even if “seasonality” isn’t specifically mentioned in the paper.
You’re right that it’s a single data point (though with a strong effect size). And, being an observational study, it simply can’t exclude all confounding factors, and will never approach the strength that a controlled study could. Nobody is disputing that. Interpretation of observational studies always needs to be tempered by the reality that it’s impossible to control for all confounders.
But within these given constraints the study essentially provides as strong evidence as could reasonably be expected (unless the study has a flaw that I’m overlooking and which, as far as I know, hasn’t yet been pointed out elsewhere). Their design of the synthetic control is statistically sound and controls for a very large number of reasonable confounders. And, contrary to the claim in the comment I’m replying to, data from other cities does not invalidate the study’s findings. As far as I can see, the study’s inadequacies are all due to external circumstances and not because it’s a bad study.
And that is why post-hoc hypotheses aren’t valid. It does not, however, invalidate this study.
Incidentally, I found somebody who prayed for all cities except Weimar and Jena, so this neutralises your anonymous friend’s prayer.
That’s not what this study says. What the study actually says is that they found wildly huge odds ratios in their confidence interval. The only “conclusion” that can be drawn from this is that the study is inconclusive.
To explain further: to positively conclude that there’s no benefit to mask wearing, the study would have had to produce a small confidence interval around zero effect. In reality, its 95% confidence interval is between 46% reduction to a 23% increase in infection. Any number between these bounds is compatible with the study results. That’s all over the place, and doesn’t allow any conclusion.
This effect is controlled for by the study, by comparing Jena to synthetic control groups.
I was able to jot down a naïve implementation and the obvious optimisation based on a lookup tables. I only vaguely remembered the Bit Twiddling treatment of the subject but, with a bit of nudging from the interviewer, I managed to implement and explain the variant that runs in O(set bits) (“Brian Kernighan's way”). I got the job.
Now, it’s fashionable to deride this this kind of code interview as unrealistic and unhelpful. But in my first week on the job, by sheer coincidence, I had to use the function. Obviously there are existing, efficient implementations, including intrinsics. But knowing how to derive an efficient implementation certainly didn’t harm. My job has since evolved into different responsibilities but low-level algorithmic knowledge is still important. I’m not sure testing for it in job interviews is generally a good idea, and designing good job interviews is certainly a big topic. But in my particular case it happened to be a relevant, fair test of my abilities.
> Maybe read the paper … before assuming that it doesn't work
I don’t assume that. In fact, I know that using ML works on many problems in genetics. What I’m less convinced by is that we can expect a breakthrough due to ML any time soon, partly because conventional techniques (including ML) already have a handle on some current problems in genetics, and because there isn’t really a specific (or flashy) hard, algorithmic problem like there is in structural biology. Rather, there’s lots of stuff where I expect to see steady incremental improvement. In fact, in Wikipedia’s list of unsolved biological problems [1] there isn’t a single one that I’d characterise specifically as a question from the field of genetics (as a geneticist, that’s slightly depressing).
But my question was even more innocent than that: I’m not even that sceptical, I’m just not aware of anything and genuinely wanted an answer. And the paper you’ve posted might provide just that, so go and do my research now.
[1] https://en.wikipedia.org/wiki/List_of_unsolved_problems_in_b...
> Genetics is amenable because it is a sequence
Not sure what you mean by that. Genetics is a field of research. The genome is a sequence. And yes, that sequence can be modelled for various purposes but without a specific purpose there’s no point in doing so (and furthermore doing so without specific purpose is trivial — e.g. via markov chains or even simpler stochastic processes — but not informative).
> There are plenty of inferences that you would want to do on genetic sequences
I’m aware (I’m in the field). But, again, I was looking for specific examples where you’d expect ML to provide breakthroughs. Because so far, the reason why ML hasn’t provided many breakthroughs in less about the lack of research and more because it’s not as suitable here as for other hard questions. For instance, polygenic risk scores (arguably the current “hotness” in the general field of genetics) can already be calculated fairly precisely using GWAS, it just requires a ton of clinical data. GWAS arguably already uses ML but, more to the point, throwing more ML at the problem won’t lead to breakthroughs because the problem isn’t compute bound or vague, it’s purely limited by data availability.
I could imagine that ML can help improve spatial resolution of single-cell expression data (once again ML is already used here) but, again, I don’t think we’ll see improvements worthy of called breakthroughs, since we’re already fairly good.
Which part of genetics are you thinking of? Much of genetics isn’t amenable to this kind of ML, because it isn’t some kind of optimisation problem. And many other parts don’t require ML because they can be modelled very closely using exact methods. ML does get used here, and sometimes to great effect (e.g. DeepVariant, which often outperforms other methods, but not by much — not because DeepVariant isn’t good, but rather because we have very efficient approximations to the exact solution).
This isn’t safe to argue at all. In fact, it’s completely wrong. We know that the poor UK response is almost exclusively due to politicking and government corruption. The scientific advisory panel of the government (SAGE) hasn’t always been right in their assessment but they very quickly produced rigorous working models and solid recommendations, most of which have mirrored (and continue to mirror) the international consensus. Furthermore, public research in the UK has, sometimes against the active opposition of the government, done stellar work to ramp up testing and genetic sequencing.
For instance, several institutes (incl. the Crick Institute in London and the University of Cambridge) had extensive testing capabilities set up in record time, but their offers to official channels were ignored for weeks, if not months (the Crick in particular simply ignored this and already provided testing internally and externally, at a time when basically no country had widespread testing yet).
Likewise, a collaboration of different institutes quickly set up genome sequencing pipelines for COVID-19 samples, and as far as I know the Sanger Institute is sequencing more COVID-19 samples than any other individual entity in the world: https://www.sanger.ac.uk/about/who-we-are/sanger-institute/t...
But that’s a completely open-ended question that really can’t be answered without referring to a primer of the underlying biology. I really don’t understand what you expect me to do here. It seems like you’re asking me to prove a negative and don’t tell me which negative to prove.
> … I am skeptical that we can say anything conclusive.
As I said we can say some things. And, contrary to your claim that I haven’t yet “named one thing that we know”, I’ve actually given a very concrete example: we can completely exclude the (often-cited, but completely unscientific) risk of viral RNA incorporating into the genome. This isn’t a straw man, it’s a frequent claim by opponents of the COVID-19 RNA vaccines. In fact, as far as I can tell this is by far the most prominent claim.
In the same vein, one can of course make long lists of potential long-term risks (cancer, Alzheimer’s, diabetes) — but unless these are plausible, this is unproductive. For most of these, there’s simply no biological connection at all. Demanding that all such far-fetched risks be rigorously excluded is unreasonable. By the same logic you could never cross the street because you can’t rigorously exclude the possibility of getting hit by a car, or a meteor. Rational risk assessment is always tempered by likelihood estimates.
The most likely, rational, long-term side effect of RNA vaccines was hypothesised to be a severe immune reaction, which might lead to the development of an autoimmune response. However, if that was the case, we would see the same effect in long-term animal trials, and we would see the start of this effect even short-term in human trials. But by now we have evidence against both of these: long-term animal testing shows no indication of an autoimmune response, and human trials don’t show any short-term ramping up of such a response.
So the most likely, hypothesised possible long-term effect is contradicted by existing evidence. Which goes back to my point: we do know some things.
Do I claim that the vaccine is risk-free? No. For conventional vaccines we have decades of data showing their safety. The evidence for brand new mechanisms obviously isn’t on the same level, and we can’t categorically exclude unknown interactions. But it’s really hard to communicate the magnitude of this risk, except to say that it’s really very small — because we can exclude plausible risks.
I’ll reply again: a lot. You can’t possibly expect me to summarise the vast, complex state of the art knowledge of RNA biology for a lay person here — it literally fills books. At the very least ask more specific questions, I’ll be happy to answer them, if I can.
> You clearly know nothing about the long term effects
Wrong. I’m no expert on all aspects of RNA vaccines, but I am an expert on RNA biology. What I do know allows me to conclusively exclude the possibility of the RNA in vaccines incorporating into the host genome (because that notion is simply not coherent). I’m not parroting any line here.
A lot more than you’ll find easily digestible in a single Wikipedia article.
> The UK government is producing AI software to "to process the expected high volume of Covid-19 vaccine Adverse Drug Reaction (ADRs)"
The volume is expected to be high not because many actual, serious averse reactions are expected but because the vaccines are expected to be given to many people at once, and every potential averse reaction will be recorded. Most records in such ADR databases are causally unrelated to the vaccine, and merely occur coincidentally; and the vast (>99%) majority of the rest are occasional mild reactions.
> But yes, label people questioning the rushed out vaccine as "conspiracy theorists".
I’m not. I’m labelling specifically those people as conspiracy theorists who make up bullshit that isn’t based on actual biology but rather on complete fiction. In other words, who spew baseless lies. I’m all for robustly criticising these vaccines. But it has to happen scientifically, and by experts.
In particular, based on our biological knowledge, we can categorically exclude claims from conspiracy theorists about the RNA likely being integrated into our genome: this just isn’t how any of this works, it’s fiction.
Interestingly that verdict also claims that URL encoding is a valid, effective encryption measure (I’m not kidding! See [2]; the German word here is “Prozentcodierung”, i.e. percent-encoding).
The court in question (LG Hamburg) is infamous in Germany for its technically illiterate, consistently laughable verdicts in IT-related cases (this isn’t a recent thing — it’s been going on for about two decades).
[1] https://en.wikipedia.org/wiki/Rolling_code [2] http://www.rechtsprechung-hamburg.de/jportal/portal/page/bsh...
> long term effects failed dramatically in animal testing.
That is incorrect.
> They use tests that can give false positives (PCR>26 cycles)
This false claim about how PCR works is a deliberate lie spread by conspiracy theorists. This is emphatically not how high cycle numbers of PCR are used.
> Or just look at symptoms that are similar to flu.
What does that mean?
> They do not have a proper (long term) placebo group (why not also use untreated patients as well).
They do that.
> Other factors with huge impact as health, habits, food, Vitamin-D.
That’s why you use large, randomised cohorts.
> Big pharma wants to collect "their" billions for their medicine.
I’m in favour of socialising big pharma companies. But this claim is still bullshit. Pharma companies can hike prices for working medication. They don’t need to invent fake medication that will be exposed in the long run and leads to company-destroying lawsuits.
That’s patently untrue. We have data, from the phase III trials. Is the data final? No. But it’s good evidence for general safety. As for long-term problems, we have good theoretical reasons (based, in turn, on experimental data) to suspect that no such risks exist. In fact, we know quite a lot about how foreign RNA behaves in cells and while there are potential mechanisms to cause issues (most importantly strong immune reactions), there are no known plausible mechanisms to cause long-term problems.
We can’t fully exclude the possibility of long-term averse effects, but we do have data supporting the vaccines’ safety (both direct and indirect, experimental data), and most experts are confident that there won’t be any such effects — confident enough to put their own health on the line: many are enrolled in the ongoing trials.
The “standard example” for natural herd immunity eradicating a disease is the black plague, but most experts now believe that herd immunity isn’t actually responsible for its disappearance from Europe.
But I guess the parent comment was using “reprogramming our bodies into vaccine factories” as a hand-waving description rather than a technically precise term. And that description is then roughly correct: with RNA vaccines, it is correct to say that our bodies are being triggered to produce the actual “vaccine” themselves; namely (at least in one type of RNA vaccine), the body’s cells are translating the injected mRNA to produce antigens, which is what a conventional vaccine contains, and which, in turn, produces an immune response.
I strongly prefer R for data science, but its dependency management story is poor, even compared to Python’s (which, in turn, is poor compared to Rust/Ruby/…).