msamwald

On the other hand the word "attention" in deep learning is often used non-intuitively, e.g. "one token attends to another token" in self attention is hardly analogous to human cognitive processes.

This is certainly a nice paper, but it is also a bit puzzling that this was noteworthy enough to be published in Nature.

But why would Libra increase rather than decrease interest in Bitcoin? Is there any actual rationale?

What is actually shocking is that these data are made so easily accessible when you are logged into Google anywhere. There should be additional hurdles to view the entirety of a person's history with Google beyond merely being logged in. The potential for very easy and all-encompassing snooping (e.g. from friends and colleagues) is staggering!

It is noteworthy that such statistics are not solely reflecting preferences, they may also reflect effects of the dating app itself! Among the many hidden psychological bugs in Tinder, an especially toxic one is that new users are presented with the most attractive profiles first. So the fact that many men don't get anything while others get a lot might be strongly amplified by a system that simply HIDES AWAY those other people.

The hypothesis is completely implausible, and it is quite irritating that it keeps popping up in discussions in the 'hacker' sphere.

There is simply no way that bicameralism would be the dominant phenotype only a few thousand years ago, and be completely absent in any present-day population, anywhere. We would still see it in isolated populations.

It is your view that is incoherent. A brain and and a PC are physical systems that work based on causal physical mechanisms. Both can run algorithms, e.g., my brain and a computer can both instantiate the same algorithm, albeit in very different ways. Your argument actually seems to be based on some notion of function, or extrinsic utility, or teleology of some sort, i.e., computers are artifacts created for a specific use, while animals and their brains just exist in themselves, or something to that effect?

Your argument is solely based on the rather odd semantics you ascribe to the term "computer". In common contemporary parlance, hardly anyone would call an abacus or a waterfall a 'computer' just because you can use such things. Conversely, I think you would not cease to call a desktop PC a computer if it is solely used for playing games and never used for 'making computations'.

When people call a thing a 'computer' they mean that it can run (potentially complex) algorithms to produce outputs based on inputs, in many cases it would also be assumed that it has memory, and it would not have much in the way of direct physical interactions with the world (although it might be part of systems that have such interactions). Based on these notions, it seems very valid to interpret brains as some kind of computer, made out of cells instead of metal and silicon, and formed by evolutionary pressure instead of human engineering.

Biologically whales are clearly mammals and not fish. Whether bony or cartilaginous fish are closer to mammals does not put that fact into question.

The article they cite claims that people who were less exposed to sun had higher disease risk [1]. As far as I saw, the cited article does not include sunscreen use in the analysis. So it might very well be that sun exposure while wearing sunscreen might have the same benefit, minus the increased skin cancer risk and accelerated skin aging.

EDIT: Besides that, there is very likely a huge confounding effect that was probably not properly corrected for. The persons in the "high sun exposure" group were much healthier than in the "avoiding sun exposure" group right from the start (Table 1)!

[1] https://onlinelibrary.wiley.com/doi/full/10.1111/joim.12496

I always found that in introduced unpragmatic, counterintuitive roadblocks for actual knowledge integration. The most annoying thing is how it carves up reality into representations of different granularity, which are then made disjoint.

I don't see how this is a competitive advantage when any company or institution can outsource these kinds of annotation jobs to low-wage countries.

The placebo group experienced adverse effects that were classified as side effects.

It does not mean necessarily mean that the effects were caused by being given the placebo ("placebo effect").

It's important to understand that if a clinical trial shows some improvement in the control group that received placebo, this does not need to be due to some placebo 'effect' at all.

The status of patients usually changes all the time. If the control group had some minor improvement in some endpoint, this might be due to wide variety of reasons. For example, the disease might just get somewhat better by itself. This does not necessarily imply that this was due to patients that were given placebo experiencing a placebo effect. They might just as well have gotten better without the placebo.

Yes, but FHIR was the result of many years that had previously been wasted with HL7 v3, which was so convoluted that at some point some people in the community gave up on it and came up with FHIR to get things done. The clock did not start ticking in 2014. HL7 already wasted a lot of time and effort, to the very real detriment of many patients around the world.

I wonder if current attempts at EHR interoperability are a case of "the perfect is the enemy of the good". Obviously the ideal solution would be one where well-structured interoperable datasets are produced and shared. This would enable very targeted analysis of patient data, and sophisticated clinical decision support functionality. However, it seems like this is still not happening, and the convoluted nature of many HL7 standards is certainly not helping.

So I've come to think that it would actually be preferable to put most of the health data into PDF-like documents, and focus on optimizing the accessibility of the data through good full-text document search that can assemble an overview by showing document snippets (i.e., allowing to get a broad cross-document overview through document previews, rather than requiring opening each document individually).

I guess some kind of especially important data, such as information on current medication or drug intolerance should really be kept in a fully structured format to enable clinical decision support, but for a lot of other information, a light-weight system based on unstructured data combined with good information retrieval tools would already improve on the status quo quite significantly.

It seems overall okay if you accept the premises that A) A centralized health record system can do substantial good by improving effectiveness and transparency of healthcare processes, and enabling medical progress through data re-use. B) Doctors won't lift a finger to do extra work if they have no selfish incentive for doing so.

No, hats usually have a completely different aesthetic and function than hair, they don't complement the shape of the face in the same way at all. Hats are also mostly worn outside, but seem improper to wear while simply sitting at a desk or lazing on a couch. Someone who would like to cover up a bald head most of the waking time is not well served with most types of hats.

So why would not everyone be bald at ever age? Why would women not be balding? Evolutionary just-so story incoming...

The one scientific text on the matter I found via a quick Google search suggests that a shaved head increases perception of dominance and confidence, but decreases attractiveness: http://opim.wharton.upenn.edu/DPlab/papers/publishedPapers/M...

Oddly, but not surprisingly, the public press over-represented those findings as also suggesting that fully bald men are perceived as more attractive: https://www.thesun.co.uk/fabulous/4469188/bald-men-perceived...

Which is quite a common thing in this kind of discourse. People are quick to suggest that being bald is really okay and nobody should care, but give the same people 100 pictures of men and they will end up rating those with full hair more attractive than those with hair loss or baldness without further reflection.

Startup idea: creating some kind of stylish, fluffy headgear that somehow replaces hair while still not trying to pose as fake hair / a toupee. This could be huge fashion market.

Well my group and others are toying with applying RNNs to permuted SMILES strings; if it ends up being better, we will find out some time this year.

Well, as far as I know the state-of-the-art results on current drug toxicity prediction benchmarks (e.g., Tox21) are held by deep neural networks; it seems like recent AI advances HAVE proven useful in that regard.

I think there is a lot to be gained by creating better systems for searching and synthesizing diverse pieces of biomedical knowledge. But indeed this is mostly a field of research, rather than something that could be directly marketed. Did you think about going (back) into academic research?

I actually think that not wanting to become sick and die are quite universal human desires. It really is the latent driver behind lots of human activity, and certainly behind most religious belief systems.

Which is not a bad thing.

Religion developed to cope with the unavoidable misery of human existence through a coordinated belief in an imaginary afterlife free of that misery.

Enlightenment, science and technology showed us that most of this misery could be alleviated by advancing our technologies and social structures, while also revealing the imaginary nature of ideas of an afterlife.

The tragic thing is that we might live in the worst time of all: religion is obsolete, yet the transformative changes possible through technology might very well stay out of reach during our lifetimes.

According to the website they use the BiDAF model, which as a single model does not produce state-of-the-art results on the SQuAD benchmark. It is ranked 27th here: https://rajpurkar.github.io/SQuAD-explorer/

A recent paper exploring various variatons of neural machine translation architectures found that LSTMs consistently outperformed GRUs in that task.

https://arxiv.org/abs/1703.03906

"[...] the keys to happiness lie in mastering the mind, and not in changing external factors in our lives."

I have been interested in mindfulness for years, but I found it very unhelpful to follow strong statements like this one. In many cases I found pro-actively dealing with problems and trying to improve my situation is eventually easier than trying to keep my mind from them.

There is a balance between being mindful and content in the present on the one hand, and being analytical/critical/discontent on the other. Of course, most people in industrialized societies could certainly profit immensely from shifting more towards the mindful state.

No, all of the medications seem to be lumped together.

That was also my first question. I looked at the original paper [1], the medication names they used to identify individuals on androgen deprivation therapy are listed in the 'eAppendix':

Leuprolide, goserelin, triptorelin, histrelin, Degarelix, flutamide, bicalutamide, nilutamide, Enzalutamide, Cyproterone acetate, Lupron, Zoladex, Trelstar, Vantas, Firmagon, Eulexin, Casodex, Nilandron, Xtandi, Nizoral, Androcur, Eligard, Cyprostat, Anandron, Flutamin, Cytomid, Cosudex, Calutide, Kalumid

So relax, finasteride does not seem to be on the list.

[1] http://jamanetwork.com/journals/jamaoncology/fullarticle/256... (probably paywalled)