Well no, but also in the US there are research facilities working with Coronaviruses [1]. I am not advocating for this theory because there is zero credible evidence for it but there are scenarios that would make this theoretically possible.
I can't find it anymore but there was an article here published a few days ago that essentially estimated whether infections with SARS-CoV-2 come from homogeneous transmission, i.e. everyone who has it infects a few people, or from highly concentrated transmission, i.e. there are a few people who infect a large number of others. The latter seems to be the case for SARS-CoV-2. This to me makes me believe 2. is correct. We know that this disease is non-lethal for the large majority of people below 40-50. It is thus not unthinkable that it was able to spread widely in those groups before finally hitting more vulnerable parts of the population all at once in March, leading to the devastating effects we observed. Additionally, in a model where everyone is equally likely to be infected, the number of deaths would reflect the percentage of people having the disease. That is why people think there is an upper limit of a few percent of infected. This however breaks down entirely if there are clusters of cases within different subpopulations.
I am sorry but this is a ridiculous thesis. This will be over eventually, either by developing a vaccine or by mutation into a more benign strain. Nobody in their right mind will practice social distancing just because they are used to it. No system that goes against basic human drives is stable. Right now there is a risk, benefit balance but once that risk is gone, what exactly will balance this? Laws? They are already in conflict with the constitution, good luck justifying any of this if there is no deadly pathogen around.
The worms are very special. Their nervous systems are weird because they have been heavily optimized to be small. A huge push to get the full connectome of Drosophila is coming to an end right now, we are not there yet fully, but close. This research has already elucidated many things about how their brains work. People are discussing how to do the mouse now. In conjunction with functional experiments this research already explained many pathways such as olfactory associative learning, mating behavior and many others. None of these understandings came from simulations, but from multiple experiments and study of selected subcircuits. There were also successes in simulation of the visual pathway of drosophila, based on the connectomic data. For example this study [1] was able to reproduce direction sensitivity of a specific set of neurons. It isn’t necessarily true that we need to fully understand the worms before we should and can move on to more complex nervous systems and successfully understand them. It might just be that neuron abstractions don’t work well in worms, because of the mentioned evolutionary pressure to optimize for size.
Agreed. The business model seems to be to give access to the rent to own deals via the sample subscription fee. Don’t think they make any money of their plugin deals. I’m also not arguing it’s too expensive or a rip off. But it’s still a large amount of money for software, in the private space at least. The rent to own thing seems like a smart tool to get rid of the barrier of entry.
The success of splice would disagree with your notion that “users hate subscriptions”. Given the horrendous price point of many of these plugins it seems to be perfect for a subscription based model. To me it always seemed there is more of a pushback from the industry producing vsts than from the consumers.
Are there no people who have to receive regular blood donations? Would you not notice such an effect in these people? I guess the donors are probably not all younger than the receiver but on average this is probably true. Does anyone know of a subpopulation that might qualify?
Masks are widely available! There isn't a single person without mask in the supermarkets. Not one. Social distancing is widely observed. I don't know where you get your information from but my experience is very different.
In addition to this, the timing of these announcements in the last two weeks were suspiciously close to lockdown relaxation dates. I would not be surprised if the RKI is using these numbers as a tool to influence behaviour, even though they know that there is a rather large error bar on the R factor. A case in point is that before the lockdown restrictions were lifted the R factor was around 0.9, but the RKI said the uncertainties are so high that we cannot be certain it is even below 1 - and suggested delaying the opening. Also, internally leaked government documents a few weeks ago showed that there is an explicit strategy of the government, to generate fear in the population in order to make people comply.
It is not possible. The number of active cases peaked 2 weeks ago and is dropping ever since. What do you suggest? Keep everything closed for a minimum of 18 months? What do we do if we never find a vaccine? It is absolutely the right move to open up carefully once your healthcare system has enough capacity.
Do you have an example from the last, say 20 years, where an actual amateur made any significant scientific contribution to an established field? I have strong doubts such a person exists. Simply because science, at the very least, requires you to be aware of what has been done already. Of course there are collaborations between researchers but this is a totally different point.
In general this is done out of an appreciation that no work exists in a vacuum. "We", takes away from the individual and attributes the work to a group of people, even if they are not coauthors. I actually think this is one of the better conventions in academia.
The parent explicitly talks about science in general. There is a tendency of some (often of those outside of science) that feel like science is getting worse over time and nothing of value is found anymore. The parent seems to echo this sentiment and gives very shallow anecdotes to make his point. Still it is the top comment as of writing. It’s just plain wrong in my opinion.
Many current problems are simply too complex to be solved by a single person, often because it requires expertise from very different fields. High energy physics, biology and neuroscience come to mind. There is not a single person who understands all aspects of CERN. Likewise there isn’t a single person who understands how to 1. Build a large scale Electron microscope to acquire high res images of an entire brain 2. Analyze hundreds of terabytes of data to make these images usable and 3. Analyze the results which would require highly specific domain knowledge of neural circuits or other biological features of the system you are looking at. My experience is opposite to yours. You are lost as a lone researcher in many fields and the best papers these days are written by enormous collaborations.
Actually there is no need for gradient decent at all if you have a linear regression problem, because you can find the min/max exactly by inverting your weight-matrix (pseudo-inverse if degenerate). Additionally, second order optimisation, i.e. using the Hessian in addition to the gradient is a well studied problem in the literature. My understanding is that it is in general not worth it because calculating the Hessian is very expensive (see e.g. https://arxiv.org/abs/2002.09018).
This is the weirdest article I’ve read in a long time. Essentially zero content, 90% of the article isn’t about what they are actually suggesting. It’s written to convince me that these are very fine people who even sold their stocks before advising the government! Wow! The poor guys read 10 Papers per day, they couldn’t even shower. And after all this work, they suggest an app and better testing! Ah, and Remdesivir. Good that they thought of that, wouldn’t know what to do otherwise I guess. Is this paid advertisement or what’s the deal?
Well that’s a bit my point. This isn’t really a thing where I grew up, but has become very normalized in the US. Building your resume certainly wasn’t on my or any of my friends mind when we were 16. To me this is insanity. Maybe it’s awesome though, what do I know.
Building your identity should not be tied to building your brand when you are 16. I’ve grown up in Europe but lived in the US for some time. What I noticed is a distinct lack of subculture in young people (college), everyone seems to be very homogenized and this is in stark contrast to what I perceived in Europe. Of course this is anecdotal but I wonder if this is a result of this crazy system that doesn’t allow missteps, doesn’t allow building your own identity, but instead forces you to be focused on security and your brand from an early age onwards as not to be crushed by college debts.
Thank you for the link. It's great that we get these numbers now. The relevant table for this discussion seems to be table 10. From this it does become clear that a majority of all people who died did not have any severe comorbidities.
However, playing devils advocate I would argue that 1) not all comorbidities are known. In this study they write: "Very severe comorbidities must have been evident within the six months prior to critical care and documented at or prior to critical care".
This is why I think pathology studies are needed. There are healthy young people who suddenly die while playing football because they have undiagnosed heart problems. Given the relatively low number of young people who die, I am not sure we can exclude this right now.
2) I could not find a breakdown of age for those people who were classified as having severe comorbidities. Even though this number is likely low, I wonder what the overlap is between severe comorbidities and those who died young.
Having said all this. I also believe that current behaviour should be conservative if there is uncertain data. Discussion should be allowed though.
Yes, there is anecdotal evidence that healthy people with high viral loads can die from this virus. It is a good null hypothesis to assume this virus could kill you. Everyone should be cautious right now. That does not mean it should be treated, and promoted, as a scientific fact to shut down discussions about the hypothesis.
There is a trend here on HN to react very cautiously and critically to studies showing “positive” results, I.e. lower apparent death rate, preliminary evidence for drug efficacy etc. This is a good thing but it should be applied in equal measure to everything. This doesn’t happen as this post shows. Where does your believe that healthy people die come from? Why is it okay to make this statement in such vigor as you do? Show me a well designed study that proves this point. I am not aware of any large scale pathology efforts pointing in this direction. Until this is unequivocally shown, statements like yours should be treated with the same reaction your parent received. It’s your opinion. Don’t state your opinion as facts.
I´ve seen plenty of them in Zurich. Not at Bahnhofstrasse of course but in Wipkingen for example they are everywhere. I had one in my backyard and there were multiple street facing ones as well.
Nothing is smart yet. What people understand under automl is different but I’m referring to meta learning here, be it the architecture, the optimizer, the learning rate etc. Optimizing these things requires an outer loop in addition to training the architecture and for architecture search in particular this boils down to training lots of models and evaluating all of them. There are ways of making this search a bit better than random search with techniques such as regularized evolution (essentially evolution with a bias for younger individuals) or other tricks but all of them require huge compute resources.
Okay, but in this case a better assumption seems to be, the stock market is a function of the economy and political interventions. Or would you suggest this function does not depend on the economy at all?
If with AutoML you mean substantial architecture search (as your question seems to suggest) then the answer is it’s feasible only for the largest players in the field. It’s possible to use some tricks but even then your standard cluster with a few hundred GPUs isn’t going to cut it for large scale problems. That said, any problem that e.g. google deems valuable enough can be meta-optimized to oblivion. Same holds if you have a huge pile of cash and think a few percent performance boost is worth it.
I’ve read through your GR part. Just to counter all the negativity here, congrats for pulling this off! I felt your papers do a much better job in explaining what’s actually the idea. Probably would be better to directly link to them here on HN. I’m a bit rusty but everything you write seems very reasonable to me. One thing I was a bit nervous about, and correct me if I’m wrong, is that you seem to somewhat artificially put in “causal invariance” by restricting all hyper graphs you consider to those that have this property. This isn’t obvious from the rule itself and it seems like it could just go wrong at one point deep into the time evolution. Are you worried about that?
Edit: Additionally the model seems very flexible. To the point that I’m unsure whether it’s surprising to be able to recover GR and other properties. I’m out of my depth here but given that string theory seems to face a similar problem of model selection have you looked at equivalences between those or does this not make sense in your eyes?