Lots of condescension here, but supposing that overnight returns are in fact on average substantially greater than intraday returns, what is the layman-friendly, non-conspiracy-theory explanation of this phenomenon?
The problem is that if you communicate to the public at the level of normal scientific certainty -- with all the methodological and statistical caveats -- it's very hard to generate the moral authority needed to push sweeping mandates on a population.
Political and scientific leaders knew this, and they made a decision to exaggerate the level of confidence they had or should have had in several of these matters. No one seriously expected leadership to have complete knowledge from day 1, but that's not the criticism. Nor is the criticism that facts change on the ground in fast-moving situations. Of course they do.
The criticism is that they knowingly overstated their factual case at the time so that they could implement their chosen strategies, even to the point of suppressing legitimate scientific dissent, and are now unconvincingly trying to use "facts on the ground change", "science learns over time", and "of course we couldn't have been expected to know everything" as excuses for those decisions.
If you're making very confident policy-guiding assertions to the public on behalf of Science (TM), and when you're right, it's evidence of how great Science is, and when you're wrong, it's because Science is a process of continual revision and uncertain information, that creates a bit of moral hazard. It works internally in science, where there are no consequences for being wrong other than wasted time, but not in the real world where there are real consequences for being wrong.
My big eye-openers (some from postdoc) were more about the sociology of science than the day-to-day productivity:
- Even the most blatantly wrong and illogical published work can only be displaced by another publication that explains/does the same phenomenon better; i.e., people are going to keep believing in phlogiston until someone shows them oxygen. If you simply point out inconsistencies in phlogiston theory, in person or in writing, they may well make a variety of unwanted psychological deductions about you.
- Similarly, nobody actually enjoys being around critics or enduring criticism, and therefore you will observe many senior scientists partially avoiding the major downsides of being a critic by artfully concealing criticisms inside what sounds to the uninitiated like mutual affirmation sessions. You have to listen very closely and learn the lingo to pick this up.
- Never question a scientific superior (other than maybe a direct mentor or very close colleague) with any other approach besides "I have a helpful suggestion about how you can maybe reach your intended destination better/faster/more precisely". Regardless of where that destination might be, such as off a cliff or into a wall.
- The opinion/fact ratio you are allowed to have as a scientist is directly and very strongly correlated with seniority, H-index, and so on.
- The incentive structure of scientific publication is such that there are big rewards for being right on an important question, bigger the earlier you are to the party, and little to no penalties for being wrong, so long as the error cannot be provably and directly linked to fraud. There are a variety of interesting consequences to this incentive structure.
In addition to ringing true, this seems largely in line with Thomas Kuhn's thesis in his Structure of Scientific Revolutions [0], a book which despite its shortcomings, should be required reading for anyone in a STEM field.
Kuhn's thesis doesn't have a lot space for the sociology of science to have this kind of influence. Of the classical theses of scientific progress, this comes closer to Lakatos' thesis of research programs.[0][1]
[1] Lakatos, Imre. (1978) The Methodology of Scientific Research Programmes: Philosophical Papers (J. Worrall & G. Currie, Eds.). Cambridge University Press.
> The incentive structure of scientific publication is such that there are big rewards for being right on an important question, bigger the earlier you are to the party, and little to no penalties for being wrong, so long as the error cannot be provably and directly linked to fraud.
This is fantastic insight and I'd like to thank you for sharing it with our group.
Would you agree that the model of rewards for correctness and penalization only in the case of fraud is the core feature of science? And what separates it from business or politics where being an honest failure is worse than being dishonest but successful?
Again, this is a great post, and I think you have a fantastic future in the sociology of science!
I think science is too big a thing to have a small set of "core features", and the question of how to usefully define "honesty" in a scientific context is another big topic, but reading about "bullshit" (the term of art that has its own literature, not the colloquialism) is a good place to start thinking about it.
I would suggest that fraud is one of the rarest types of dishonesty, because people who are both smart and dishonest have less risky ways to proceed, and that such people are very glad fraud exists, because it misdirects attention away from their arguably more damaging and prevalent methods. Feynman has a passage about how honesty in science is more a state of mind, which I agree with. But really, the techniques to be dishonest with low risk are the same in science, journalism, politics, and business.
My field isn't sociology of science though; these are just views from the genomics trenches.
I'm joking a bit with the style and my limited experience leads me to agree with the OP.
The middle paragraph includes my sincere response that a system where discovery is rewarded, failure forgiven, and dishonesty punished is ideally suited to the mission of science.
So I was left wondering if the OP would expand on what their thoughts were about the interesting consequences.
Very early on, I noticed that graduate students tend to be idealistic, postdocs extremely cynical, and faculty ruthlessly pragmatic perhaps to the point of occasional shortsightedness. Clearly, something about this progression is expected and normal. I'm a postdoc now, so I'm right on schedule.
I think the way it ultimately works is that you have to be disillusioned from the grade-school fairy tales told to the public about how science works before you can learn to live and work in the environment that actually exists rather than the one you wish existed.
> "Never question a scientific superior?" Not parsing that concept, please elaborate.
tech < grad student < postdoc < junior faculty < full prof < Big Guy/Gal < Nobel Laureate < NIH Director
People above you in that chain will accept limited feedback on methods to attain their chosen goals and will greatly resent questions about whether their selected goals are worthwhile/realistic/rational, or whether their gestalt vision of the field's conventional wisdom is correct.
"People above you in that chain will accept limited feedback on methods to attain their chosen goals and will greatly resent questions about whether their selected goals are worthwhile/realistic/rational, or whether their gestalt vision of the field's conventional wisdom is correct."
Corporate management has the exact same situation.
Yeah, in corporate world, at least you're paid to not care, and can change jobs easily. In academia, you're paid shit, and changing labs is not nearly as easy.
I agree with most of the GP's points and I don't think of them as defeatist, but rather a call for realism when dealing with people (versus data, which have no ego to bruise). It's very hard to devise a system that rewards individual achievement without ever falling prey to classic human flaws. The good news is that science over time tends to be self-correcting, and all that requires is a commitment to shared principles and methods, combined with enough anarchy that no one individual can screw up an entire field (Trofim Lysenko being the most extreme example, but any bureaucracy can accomplish this).
The presentation is cynical, as xab31 themself attests, but I don't think it's defeatist.
1. Bad work only being displaced by good work: everything works like this. To replace some useless commercial product (take your pick) someone has to come up with something better. Same goes for information.
2. Nobody liking criticism can be rephrased as it being important to attack ideas, not people, when you have to work with those people.
3. "Never question a scientific superior" is the first piece of advise I think is too cynical. As a warning against undermining a colleague in public when you need their support, I agree, and that's kind of a restatement of #1 and #2. But science really does have a culture of publicly debating contentious ideas. You can definitely be more critical in an event specifically held as a debate / open forum than in a presentation Q&A though, and at a social event it's polite to be at least vaguely supportive.
Kind of a tangent to the later points: Day to day scientific research is mostly chasing dead ends and other activity that is (in hindsight) mostly useless, but there is genuine societal value in having a large body of skilled workers available. That is, science spends a lot of time spinning its wheels trying to figure out the right question to ask, and once this becomes clear there is rapid progress. This means the papers published in between the breakthrough periods aren't really worth paying attention to unless you work in that area. Having a lot of scientists and engineers in the workforce so we collectively have a decent chance at obtaining and exploiting next breakthrough is the point, the papers are just a byproduct.
>"Never question a scientific superior?" Not parsing that concept, please elaborate.
If you think you've been put on a bum topic or your supervisor has put you on the scientific equivalent of a PIP with no way up or out your room for maneuvering is limited, to put it politely.
I understand the impulse to not want to be defeatist, but sometimes it’s both easier and more productive to stop running into the same walls over and over and instead find the path around them.
Well, as defeatist as going to work fo a FAANG and not expecting that your managers will give a toss about fairness, their users privacy, or the spirit of regulations. Life is like this. Right now in the African Savannah a lion is mauling a gazelle, it happens daily.
This is true here, as well. I once asked about something related to SSL/TLS (fairly politely) and was kind of mockingly escorted by some groupies to the corner since I apparently responded to an Apache developer.
I was just trying to learn. Learning bad is what I learned.
The sociology of science is so interesting. (Not the field, but the subject.) Here are some of my favorites quotes/thoughts:
This one is a direct contrast to your advice (which speaks volumes about what's wrong with academia): "A good scientist, in other words, does not merely ignore conventional wisdom, but makes a special effort to break it. Scientists go looking for trouble."[0]
This was written about physics at Caltech, but applies more broadly. It explains why the ability to 'manage up' is so critical for early-career success. "[...] departments are run, for better and worse, by the professors who often lack managerial experience. Worse, they are generally unaware of this shortcoming, assuming incorrectly that management is trivially easy compared to their topics of study and merits minimal effort. We have now seen the consequences of this lack of attention." [1]
Academic politics is a great reason not to stay in academia: "Look for environments where competitors see themselves as playing a game, rather than fighting for survival — this prevents rankings within the hierarchy from becoming an existential problem." [2]
This book has a great chapter of career advice, here's a gem: "Don't build a pyramid. Everyone seems to build one pyramid per career. A pyramid is an ambitious system that one person really cares about and that winds up working well, but then just sits in the desert because nobody else cares the same way. This happens usually just after leaving graduate school." [3]
"In general, status-conscious places are miserable for everyone, and the more, the worse." [3, next page]
Gatekeeping is predictable from the incentive structure: "For all the high-level talk about how we need to plug the leaks in our STEM education pipeline, not only are we not plugging the holes, we're proud of how fast the pipeline is leaking." [4]
"So why am I not an academic? There are many factors, and starting Tarsnap is certainly one; but most of them can be summarized as 'academia is a lousy place to do novel research'." [5]
"...whereas Newton could say, 'If I have seen a little farther than others, it is because I have stood on the shoulders of giants,' I am forced to say, 'Today we stand on each other's feet.'" [6]
It is several repeated and very costly attempts that I made to do just that which leads me to give the advice I did.
The pyramid quote is an interesting one. Obviously there is a tension between being passionate about an idea/goal/cause but not being overly siloed. It seems the best-case scenario is: pick your passion, find some people who're thinking in the same general direction, and compromise the vision among yourselves.
Let's just say that the thought of solving some of the problems I'm interested in from outside academia has occurred to me. But I'm sure it's not all sunshine and rainbows on the outside, either, and moving from academia whose primary motivator is risk aversion to something like a startup is an extreme culture shock, the more so because my objective would be building something real, rather than bilking gullible VCs into an acquihire.
Well, I do aging research (mostly from a computational+biochemical perspective). I've met most/all of the important players in the field, and it baffles me how this important area of research continues to be a backwater, as far as the public's concerned.
It's hard for me personally to think of something more important than aging, so if I were to expand outwards, it would be to pursue the same goal, but maybe with fewer constraints. In general, I'd work towards streamlining and automating certain aspects of it. Technologically, the field is in the Dark Ages. There are realistically ~200-300 (max: 5000 including subordinates and techs) people in the entire world working on this seriously, which is fairly mind-boggling, considering that it is the primary risk factor for cardiovascular disease, cancer, and indeed COVID-19, along with many other diseases and the more transhumanist and futurist implications.
Young people need to realize that the things we love about science: the uncompromising search for the truth, its international and no-boundaries character, the ability to bow down to evidence, the ambition of the ideas, are just a very distilled fraction (basically the highlights) of a what it is a very mundane, fragile, political human activity, full of petty and lame characters, absurd situations and pathetic developments.
I work in an adjacent area and agree this is all good advice.
OP, how did you even get the sequence to begin with? I have a friend who has an immunodeficiency which is almost certainly due to a rare genetic disorder and want to do a very similar thing. Despite contacting his physician, fellow researchers, and even my institution's president -- with friend's full cooperation -- no one is willing to pay for it.
I'm at my wit's end to the point that I'm starting to think the only viable option is paying for it out of pocket, but it's not cheap.
A question you might want to ponder is: suppose you isolate the problem to a single missense/nonsense/truncation mutation in a protein that seems likely to cause the phenotype. How do you plan to use that information? In theory, there is gene therapy, but in reality, given how much effort I have had to go through just to get this fellow sequenced -- and I'm a PhD working in genomics with a lot of contacts -- creating a custom one-off gene therapy solution seems like it would be a very tremendous undertaking.
There is a very difficult problem here in that rare or "personalized" disease treatments are: A) not profitable, so drug companies have no interest, B) there are mountains of paperwork, IRBs, consent waivers, etc, involved in developing an experimental therapeutic, C) by definition you cannot do a proper clinical trial on a one-off, and D) it requires several different types of expertise to pull such a thing off. Sadly this means that it almost never happens, even though I suspect there are a lot of severe and lifelong genetic disorders which could be diagnosed and treated with technology available today.
Based on my experience so far, I suspect that even if you were to hand his physician very strong evidence that "the problem is caused by this specific single mutation", the response will be "OK, thanks". You should not make strong assumptions about them being able to take it from there. All this is based on the best-case scenario of it being a single variant in a coding region; if the disorder is caused by multiple variants at different loci, anything you find will probably not be actionable.
> OP, how did you even get the sequence to begin with? [...] no one is willing to pay for it.
My neurologist ordered sequencing for me from Invitae, to determine the subtype of Ehlers-Danlos I have and rule out neuromuscular diseases. She said insurance usually covers it, and it's only a few hundred bucks if they don't. Invitae appears to do WGS for such panels. I've also heard of Nebula genomics offering affordable WGS and exome sequencing.
She said she'd take a look at the results, and if anything popped out as unusual, I'd see a geneticist.
> A question you might want to ponder is: suppose you isolate the problem to a single missense/nonsense/truncation mutation in a protein that seems likely to cause the phenotype. How do you plan to use that information?
Identify the molecular pathway involved and see if there's any drugs available that might modulate it in a therapeutic way. You might also identify similar diseases that might share similar treatments, once you know the etiology.
Once a mutation or gene responsible is identified, other patients can be as well, which can slowly lead to mouse models and clinical trials etc.
This. It's not my field but I strongly suspect I have a troublesome mutation of some sort. If I was able to track it down what good would that do? (I almost certainly inherited it which is why I think it's genetic.) Sequencing isn't that expensive these days, I keep thinking about it but even if I found enough others to pin down the mutation what good would that do? "Success" would simply be an accurate diagnosis, nothing more.
Not OP, but I got my whole genome sequenced from sequencing.com. It is currently around 399 USD. The sequencing is done by Nebula Genomics, nebula.org, where it is currently 299 USD.
You get to keep all the raw data from the sequencing and you will get some reports included.
I did it in order to do a genetic disease screening as I was very sick with strange symptoms for a long time.
Thanks to you, and others, for sharing. I hadn't yet resorted to looking in the consumer space. In research (and presumably clinical)-land, the costs are substantially higher.
I'll be looking into it further to figure out whether there is some tradeoff here, or if it is just typical cost bloat for medicine/academia.
I haven’t done very thorough research, but my impression is that there is a considerable cost bloat if you go through hospitals and similar, but that the genome sequencing is essentially the same.
So, we have a climate crisis, uncontrolled health care and college costs, decades of pointless war, mass incarceration, and a variety of other crises that cause a lot of death and suffering.
Yet isn't it strange that the one crisis we have chosen to pull out all the societal stops for, to radically reorient all of society and put it in stasis for, is COVID-19. Odd coincidence that most of the first list affects young people, COVID-19 primarily affects old people, and the political leadership of the developed world happens to be comprised of old people.
And the subset of old people responsible for handling the pandemic hasn't even managed to do that properly. This is the same category of people who hollowed out unions, induced globalization, and generally kicked out the ladder beneath them in a variety of ways.
I do care about my parents/grandparents, I've been vaccinated, and I'd wear a mask around an old person. That's the absolute maximum I'm willing to do voluntarily and feel fine about that. I'd even venture to suggest that if someone has a problem and demands that all of society radically realign itself to fix/prevent it in a way that's disproportionate to society's other needs, it's not society that's being selfish.
One of our graduate students is thrilled about this paper, although he tends to do that with any new CS advance that seems sensational and that we barely understand (we do bioinformatics). He said that it stood to reason that if it can mmult 100GB/s/core, then we could matrix multiply 12TB in a minute!
Could you translate into practitioner-level language what are the practical limitations of this method; specifically, what would the error rates induced by approximation be under some practical scenarios, when would it make sense and not make sense to use it, etc? There is a complex equation in the paper describing the theoretical error bounds, but I have no idea whether in some practical scenario multiplying some normally distributed variables, whether that would mean a 0.1%, 1%, 5%, 10% error.
Personally I think it only makes sense to use this kind of method in some real-time algorithm where speed is of the essence, the downstream results of the mmult are themselves used in some other approximation (like many ML applications), and emphatically not to make the process of drawing biological conclusions from painstakingly derived data a few minutes faster for the analyst.
I fear that you have made an impressive, but dangerous, tool to people who don't know what they're doing.
I think we only claim to be able to preprocess a matrix at "up to" 100GB/s/core. The overall matrix product will take longer and depend on the matrix shapes.
To simplify Section 1.1, we help when:
1) You need to perform a matrix product more quickly and can tolerate approximation error
2) You have a training set for the larger matrix
3) The smaller matrix is either a) fixed or b) skinny relative to how tall the larger matrix is.
Re: "an impressive, but dangerous, tool to people who don't know what they're doing."
I believe you are overestimating the usability of my code :). But more seriously, I suspect that people attempting to use our method in contexts it wasn't designed for will quickly discover that they either can't actually call the API the way they wanted to, or that the method is no faster for their purposes. We also characterize our method at least as thoroughly as any approximate matrix multiplication algorithm I'm aware of, and have a variety of (admittedly loose) theoretical guarantees. So I hope that at least those who thoroughly read the paper will have a clear idea of what it can do. Overall, I guess my current thinking is that 1) I'm not sure how to introduce a method any more responsibly, but 2) if I can be of help in ensuring that it gets used well, feel free to reach out.
It was a very effective response from Jeremy. He clearly describes his side of the story and convincingly explains why he thought he did nothing wrong.
What is a little more troubling is that he (A) describes codes of conduct in detail and repeatedly affirms his allegiance to the general idea behind CoCs, including describing "previous sexual assault allegations" as "behaviors" he "strongly agrees" should be stopped. (B) He goes some way towards portraying himself as a victim, describing in graphic detail his lack of "emotional resilience". I wonder if he realized what a thin line he was walking with (A), because he now has a "previous CoC violation allegation" on his permanent record, regardless of his acquittal.
In other words, he very strongly backs the spirit, letter, and zeitgeist of CoCs before describing why he thinks he did not violate them in this specific case. I think these were very wise tactical moves, although they leave a bad taste in my mouth, and I can't help but wonder if this approach affected the outcome. I don't recall seeing an apology ever given, let alone changes instituted, for CoC violation allegations.
My gut reaction if I were Jeremy would be to say that "This is absolutely absurd; there is no way that claiming 'X's opinion is wrong' is in itself a CoC violation, and if it somehow is, then the CoC is ridiculous. The fact that such an accusation was even able to be made under the CoC and not immediately dismissed makes CoCs look ridiculous. Why are you wasting everyone's time?"
He goes a very, very different route. I think that different route gave him a better outcome. I think it is worth thinking about why that is. We are not talking about an accusation of sexual harassment or assault here, we are talking about an almost textbook case of how CoC accusations can be overextended to absurdity. Yet he chooses to defend CoCs.
One possible explanation of this is that he is trying to draw a distinction between things that should be "legitimate" CoC violations, and his case. Another possible explanation is that he is trying to say that "Look, I strongly believe that <murder should be illegal>, really strongly. That makes me one of the good guys. Would a good guy commit murder?" The latter take is more cynical, but I think it would be more effective than my gut reaction.
I think you are vastly overestimating the significance behind having a "previous CoC violation allegation on [one's] permanent record". Who's keeping, and enforcing on, permanent records of this sort? Every org is going to call this one for themselves, and the facts are out there.
The thing that seemed to me oddest about all this is that Jeremy says the committee spoke to two reporters before hearing his side of the story, and before issuing a judgment. It is not totally clear if they were speaking to those two reporters about his case.
I would imagine that an obvious way for CoC committees to minimize their own liability would be to at least avoid talking to reporters before they have issued a judgment. Really, the process should be totally private until a guilty judgment has been made, and if the person is declared innocent, then the whole thing should disappear without ever having been made public.
> The issue, amazingly, is Fox News and it’s ilk. Yet, the conversation here is the NYT.
That is because practically everyone on HN agrees Fox is bad, biased, etc. Therefore the debate is going to be implicitly about how bad the NYT is in relation to Fox.
A related factor is that it's hard for an educated person to get suckered by Fox. There are too many garish infographics and obvious nutjobs. It just does not give even a superficial impression of being Legitimate and Unbiased and Supported by the Best Experts. But the NYT does, and that's what makes it more dangerous.
If I go into "Uncle Cletus's Homeopathy Clinick", I kind of deserve whatever I get. But if another con man has a convincingly faked (or even real) Harvard M.D., then sets about poisoning lots of people through incompetence and apathy and greed, then everyone insists it can't possibly be his fault because he has an M.D. from Harvard...
...you can see why "Uncle Cletus is the real problem here" can seem nonresponsive. It is not even especially obvious to me which is "worse", "Uncle Cletus" or Fake M.D., even if we grant that Fake M.D. is somewhat better at medicine. I know I personally could get suckered by the latter but not the former, making the latter more dangerous to me.
> A related factor is that it's hard for an educated person to get suckered by Fox. There are too many garish infographics and obvious nutjobs. It just does not give even a superficial impression of being Legitimate and Unbiased and Supported by the Best Experts. But the NYT does, and that's what makes it more dangerous.
So, same reason why scam emails are rife with spelling errors.
They might actually face more criticism if they don't turn away the part of the audience with half a brain first.
It's still used quite heavily in research also (medical is what I'm familiar with). NCBI puts a lot of critical data on FTP, as do individual researchers.
In the case of NCBI, the actual transfer protocol usually used is Aspera, but you use the FTP server to find what files you are going to download using Aspera (which is CLI only AFAIK). Thus in this workflow, browser support for FTP is actually important.
I would say that the reason it hasn't changed is that it does its job fine. It serves files. It's one of the simplest ways to do so, and I lament that in my sphere the biggest apparent replacement is Dropbox. I agree with one comment above that "legacy" code for FAANG is apparently "features which we can't monetize".
Interesting because in the original post, he declares his resolve to quit HN and declares that he's not visited it in one whole month. "Yeah, right", I thought, and indeed in the intervening 3 years, he came back.
I'd wager most of us have needed a break at some point. I myself have gone through at least 4 accounts since 2006. I periodically "delete" them by setting minaway to 999999999. The act of starting over with a new account with zero fake internet points does help keep their value in perspective, though.
