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?
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.
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.
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.
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.
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.
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.
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 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.
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.