timkam

I somewhat agree, but many traditional publishers have an optional "pay for OA" model and more and more funding providers require OA. For journal publications, we (in the EU) almost always have to pay. However, MDPI is particularly aggressive in its editorial practices and marketing tactics. In my community, we sometimes publish "smaller" works there: when early-career researchers get to guest edit a special issue in an MDPI journal and we want to support them; i.e., we sometimes fall for MDPI's marketing tactics ;-)

It's best to have journals that are owned by the community, but these take time to grow into venues that administrators respect: we can publish some work there, but it's not enough for building a competitive academic CV.

Considering this article: it's just a comment, and to me it looks like something an expert can easily tear apart. A well-established researcher wouldn't publish such a comment in an MDPI journal (as one of the other comments suggests).

Interesting. This reminds me of the story of Siraj Raval, a formerly highly influential YouTuber on machine learning and blockchain topics: https://www.plagiarismtoday.com/2019/10/16/why-siraj-ravals-...

Some years ago, when I started getting into machine learning, I couldn't tell he is a fraud. Considering that just before he was broadly exposed as a fraudster, he was invited by the European Space Agency (ESA) to give a talk, I was in good company. I think the only reason he got exposed is because he started making stupidly bold claims, attributing high-profile original research to himself. If it weren't for this, he would probably still be accepted by the mainstream because the case a few experts could make against him wouldn't be strong enough.

My point is: it is hard to see what is fraud and what isn't. If the ESA can't tell the difference, how can average parents?

I agree. Covid-related work that is disseminated in "non-medical" venues presents most likely toy examples, which is okay as long as the authors clearly state the limitations and do not try to push their research directly to the public without getting medical experts involved.

Researchers with really strong results will for sure publish in medical journals, which typically implies a massive push in prestige (impresses the administration and funding providers).

I agree and I really don't understand how mature researchers can engage in this type of guesswork. If we are in a simulation, our observed universe is also a simulation. Even if the real world revealed itself to us Matrix style, this "revelation" could also be a simulation. There is no way to tell. We can not falsify either of the contradicting statements "we live in a simulation" / "we do not live in a simulation".

(Edit: Technically, I don't agree with your "application of probabilities to non-discrete systems" statement, but I think we get the intended meaning here.)

Am I the only one who finds it strange that a preprint gets that kind of news coverage and the authors of the preprint are the only ones who are interviewed? Even if it wasn't a preprint, I suppose it's good practice to ask independent experts for a second and third opinion. (Edit: a problem I see with such news coverage is that it really encourages researchers to rush their work out and make headline-grabbing claims even before they have any idea of what the scientific community thinks about the work.)

Humans are social beings. An abstract concept that is named after its human creator establishes an immediate link to this person. My guess is that many people like this, because it adds a social dimension to otherwise "dry", technical papers.

I fully agree. As I understand it, the user was a teenager who did not know better and thought they made a sincere effort to contribute. Young and/or inexperienced contributors are prone to make somewhat similar mistakes, albeit usually less extreme. What message would it send if they were punished for this?

What worsens the problem is that publicly funded research often solves toy application problems (at least in CS); academics know that they need to "apply" their research to get funding/citations/publications with relatively little effort, but from an industry/real-world perspective, many application scenarios hardly make sense and are hand-wavy-ly evaluated (for example, in applications of CS, too often the code is not shared so that the results are practically not reproducible). The current system simply rewards researchers who manage to get into the top venues with the least effort possible, and for many who aren't brilliant the pseudo-application approach is the way to go. IMHO, this broadens the gap between academia and industry, because it's hard for a practitioner to pick out the few relevant nuggets in a stream of half-baked applied research.

These insights may be trivial, but still important to experience first hand and early in life (let me highlight that I was never an excellent or successful musician, though):

* Having solid basic skills in a profession is crucial; in contrast, almost nobody really needs a virtuoso. Practically, nobody needs a rock star or diva.

* Always be reliable, even if others are not (don't pay back unreliability by being not reliable yourself). Be the one who is on time and does his homework. (In contrast, in school I learned that what technicality qualifies as homework often does not pay off.)

* Continuity is key to get a decent chance to succeed (again, in contrast school taught me that you can make things up as you go along, as long as you are somewhat talented).

To relate to the topic: the problem with many video games is that reliability and continuity come effortless. You simply need to go online and play. In contrast, as an amateur musician, I had to make sure I have the right gear, plan the transport without owning a car, make sure I can get into a rehearsal room regularly, find the right people to play with (the ones who have the connections that are needed to actually get a gig) et cetera.

Most people who like books, films and TV, consume these media types passively, i.e., they are less immersive and hence perhaps less dangerous. With _creating_ music, I'd argue the situation is similar to martial arts: it requires discipline, and actually social skills if you want to be relevant. The same is true for many "e-sports". However, I agree with Rogan more or less, because there are just too many games that are too dull and too many young males who waste their time with them. Based on my personal experience, I would say most of my time spent gaming as a teenager was wasted, but playing music actually taught me important lessons for life. For me the social aspects, as well as the "mastery" aspects of gaming were just too simple, too simulated.

Marvel's "Knull" is a fun example that the website is actually "catching": https://www.svt.se/kultur/starka-reaktioner-pa-nya-marvelsku...

See: https://en.wikipedia.org/wiki/Knull_(comics)#Reception

No longer a Software Engineer, but still in "IT" in the broader sense.

* Left SE job for a PhD in CS, roughly 50% of a good SE salary here in Northern Europe

* Do some Product Management/Innovation advisory on the side

I left SE because I found the type of programming one typically does, as well as the Jira ticket-pushing, uninspiring. I still write code, but (almost) only because I enjoy it (a PhD in CS does not necessarily involve much, or any, programming).

Can anyone point to a more nuanced perspective on this? Ideally not only with regards to OpenAI; it is still quite common that research published in some of the "top" venues does not disclose the underlying source code.

Anecdote: I once pushed researchers to publish their code when I was peer-reviewing and it turned out that the code was super slow in comparison to the benchmark algorithms they compared with (comparison only included accuracy etc, not speed), something that I am sure the authors were aware of, but chose to not report in the paper.

This certainly depends on what the code does and on the use case, i.e. on how critical reliability and performance are. 10k code in a simple web application prototype that will be trashed in 3 months is not "a lot", because it's not critical and piling up technical dept is not a problem. 10k code to specify the core algorithms in a mission-critical, long-living system is "a lot", or at least "much more", because it needs to be crafted and maintained with immense care.

I know it's just an example in the article, but the study about real-life likeness of social networks in fiction literature seems to miss a point. Wouldn't a good writer in many cases abstract away some of the complexity of real-life social networks?

I agree, but both from a technical and a social perspective, a) Norway is in a stronger position to mitigate the economic impact of the crisis and b) Norwegian politicians will have a harder time to spin up an austerity narrative.

There are additional relevant aspects:

1. A main aspect of the response, i.e., the issuing of "recommendations" instead the implementation of outright restrictions seem to have been caused by legal concerns, which means this is probably not something that Tegnell and his team decided.

2. Norway can rely on its natural resources and on the resulting "oil fund" to mitigate an economic crisis, Sweden lacks such resources, at least to the Norwegian extent.

3. The Norwegian government is "right-wing" from a Scandinavian perspective. Many of its policies would be considered socialist in the US.

4. Most deaths in Sweden are related to elderly care and the lack of preemptive efforts to prevent covid-19 spread in elderly care homes, which was certainly a big mistake, but somewhat unrelated to many of the measures that Sweden is scrutinized for.

I think we can learn from the OS community, but this requires changing the incentive system of academia. Two ideas:

* Research papers should be living documents that are properly versioned; updates should get reviewed again. This obviously means that incremental improvements to papers and reviewing need to result in something tangible for one's CV. The versioning makes sure one can still point to a specific revision of a paper. Of course, updates should consider the implications of changes carefully (as is the case for software). For example, I might get familiar with a paper over the course of a month, and then use it as the cornerstone of one of my own works. Then, I want that the paper is stable and does not constantly mess with important formal definitions et cetera. Otherwise, people won't be able to agree on the implications the paper has for the community.

* It might be better to make review and reviewers fully visible and move to a more interactive review process. For obvious reasons, practically "double-blind" review does not work, because if you're a visible member of your community and submit a paper, most people can guess that it's yours. If we would move from "one-shot" (or: one-shot plus rebuttal) reviews to more interactive, public, non-anonymous discussions, we would get rid of at least some of the social back-channeling that is only available to the bigshots in the community.

Of course, I don't think we will get there quickly.

The same statement should apply to research papers; I am happy that the EU is pushing for this as well, with considerable success.

Masks directly help prevent spread, the app does not. It's essentially a decision-support system, and I agree with Schneier that on an individual contact tracing-level, it's probably not a good one. But _perhaps_ it helps if it's well-designed and presents aggregates of probabilistic information well. If its underlying models are good it can, for example, inform a user where not to go because the infection risk is too high, after it has gathered data of frequent close encounters between otherwise "distanced" people in a location.

In many experimental sciences, a lot of research projects now require an effort that a few authors cannot handle. There are some discussions if this means that instead of being creative, people are merely scaling up, which is probably true in some cases, but often strong empirical results that enable "real-world" progress require such efforts. But that's not Math. I work (partly) in a field of theoretical CS which is Math-heavy and attracts mathematicians and there I also feel the indirect effects of Covid-19. As a junior in the field, I need to rely on the feedback of more senior people, and these people move fastest on the whiteboard/with pen and paper, in physical meetings. The good thing about these collaborations is that we juniors can provide fresh ideas and the seniors can shoot down the bad ones quickly (before a long, formal review process) and make sure the good ones are executed with the necessary rigor.

(Note: I find the comment of "your" reviewer strange as well.)

It's hard to disagree with the statements you make in this comment, but this does not answer my question. To say "free software is important" is one thing; another thing is to say "if a GPL-licensed software includes a closed-source dependency, this is a news-breaking problem". Considering the deep dependency trees many applications have, I would assume incompatible licenses in an app or library, and yes--even closed source dependencies in GPL-licensed applications--is a problem that occurs rather frequently. I think there is a generic problem at the intersection of license management and dependency management. COVIDSafe is just an example of this, and given the prevalence of these issues, I think it is in itself not particularly newsworthy.

I would assume many GPL-licensed applications are GPL non-compliant. From a user's perspective, what are the most relevant problems when this is the case? (Excluding problems of philosophical nature that only free software geeks care about. No intention to be disrespectful; but the more practical the problems, the more likely it is that ordinary users can relate to them.)

Observations from overseas (Sweden): possibly because college/university is typically (almost) free of cost and because of Sweden's culture of continuous education, there is an increase in university enrollment applications: https://sverigesradio.se/sida/artikel.aspx. I must say that I like that in times of high unemployment, the "free education" feature allows people to further educate themselves when there are little employment opportunities. IMO, this feature takes the risk burden (that the article is mentioning) off the individual.

Isn't this the report of the case study that was discussed with much controversy in German media, in particular because the researchers hired a high-profile marketing agency for promotion and held a press conference together with the prime minister of the federal state (NRW) before subjecting the findings to the scrutiny of the scientific community?

Edit: link to newspaper article (German): https://www.sueddeutsche.de/gesundheit/heinsberg-studie-coro... Note that Gangelt is a municipality in the Heinsberg district, so the article and the report refer to the same study by different names.

What I get from this article is that OpenAI is in many aspects similar to other research organizations, but with more resources and better PR. I think this article is reassuring for anyone doing AI research outside of the big labs. However, I also think that nothing that the article reports puts OpenAI in a particularly bad light; it's just the ordinary problems a research organization typically has.

(Note: GPT-2 communications were unfortunate and insincere, but I've witnessed similar, less high-profile communication blunders.)

An observation from inside academia: there are some (potentially many) researchers who use the opportunity to push obscure models and simulations of their research sub-subdomains, even if they have no knowledge about public health, epidemiology, or even strong past real-world success stories of the methods they are advertising. Sometimes, this happens with a "talk to the media first" approach (before subjection of their work to the scrutiny of actual domain experts or even any form of peer review), which is certainly dangerous.

I agree, not necessarily because publishing in a journal one is involved in is uncommon, but rather because putting one's work there although this work is disputed in the community clearly signals that one has not managed to build consensus and does not want/is not able to fight the good fight. If it was an incremental addition to well-established work that hardly anyone in the community can dispute, it would be a different story.

Right, so I suppose that the numerous gateways zoom and others need to offer are an additional problem which implies that these services have to do "something fishy" on the signaling server. Generally, WebRTC traffic does not need to go through a centralized sever, though. It's peer-to-peer after signaling if possible, and if not it can use routing servers that merely route encrypted traffic. So I am wondering if these providers largely make use of standard WebRTC infrastructure plus gateways and how much proprietary magic they have on top/as an alternative. Of course, in no scenario the traffic is really secure.

I guess Zoom says they're end-to-end encrypted because they're using WebRTC, which probably means traffic is end-to-end encrypted after signaling, but users need to trust that zoom's signaling server doesn't do anything fishy.

Edit: I do not understand the reason for the downvotes. I am not defending the practice but am just describing their potential line of explanation. Please let me know explicitly if my comment is technically incorrect. Also, I would be interested what other vendors claim, who probably use similar technology under the hood.