sqrt17

Hetzner provides exactly that: a VPS. They're also cheaper than DigitalOcean, I assume, because EU web hosting is way more competitive and US VPS offerings have traditionally been way overpriced.

Let's also be realistic here: CloudFront is a good solution if you need a file to be served over CDN with low latency in multiple regions. Raw s3 or just a web server is much cheaper if you don't need low latency over multiple regions.

The difference being that BTC has no utility that people depend on. They can refuse to buy the BTC from the now-GenZer or even the Winklevoss twins if they'd rather have food or shelter or a chanel bag rather than BTC. The real world phenomenon that this corresponds to is consumer price inflation or asset price inflation (assuming your asset is production capacity and the medium of exchange is BTC) Being a landlord, or being an oil baron in an economy that is dependent on petrol, is different because people actually need those things. So antidemocratic aristocrats who inherited a land with substantial oil reserves can now buy lots of startups via softbank or build artificial cities because our cheapest way to acquire petrol is still to dig it out of the ground in those countries where reserves accumulated in the last millions of years, and pay local rulers for the permission to do it.

At our (fairly large) company, you can query by team (and maybe job role) but it will hide responses where the sample size is smaller than a set number (I think 8 or 10).

So yes it can be done but people have to actually care about it.

The cautionary tale about k-anonymity (from Aaron Schwartz's book I think) is when the behavior of aggregates is also something that should be kept privates - the example was that the morning run of an army base in a foreign country was revealed because enough people did this with their smartwatches on that it formed a neat cluster.

It takes substantial effort to build a good dataset, proportionally more if it gets bigger, and people like big datasets because you can train more powerful models from them. So I am not surprised that people tend to gravitate towards datasets made by well-funded institutions.

The alternative is either a small dataset that people heavily overfit (eg the MUC6 corpus that was heavily used for coreference at some point where people cared more about getting high numbers than useful results) or things like the Universal Dependencies corpus which are provided by a large consortium of smaller institutions

Probably i don't watch any of the YouTubers you're watching either - the long tail is exactly why people watch YouTube (and maybe a handful of 10k-1M subscriber channels) rather than, or in addition to, TV programme and streamable shows that are modeled after it that are less niche.

The main point of that list was: there are multiple platforms that try to be a "YouTube for people who don't like YouTube, for a point in time when YouTube gets horrible", with some people creating content there, and e.g. odysee lets creators sync their videos so you get exactly the same videos from them on YT and on Odysee. But for most creators this is still additional work for little benefit since their main audience is on YT.

Some of the more well-known YouTube creators are also on

- Curiosity Stream/Nebula (Adam Neely, Mary Spender, Ali Abdaal, Thomas Frank, Super Bunnyhop)

- LBRY/odysee (DistroTube)

- FloatPlane (Linus Tech Tips)

None of these has a nice recommender like YouTube, so you'll likely either have manual curation (Nebula, Floatplane) or drown in a sea of bad/uninteresting content (LBRY)

False dichotomy. What you call "not studying" would vary between effectively dropping out (for the people who are only at uni to get drunk and end up with a degree at the end), gaining a more thorough holistic understand of the material, or diving really deep on some of the aspects that are most interesting at the cost of others.

The drink-and-get-a-certificate people will indeed profit from being encouraged to study, but the controversial opinion here would be that university should only or mainly care about those.

Daimler (Mercedes) is quite close, and they complained loudly when Boris Palmer decided to drive a Prius rather than a Daimler car when he became mayor of Tübingen.

So yes, the auto lobby is quite powerful but the extent to which they can interfere with the business of a small liberal university town is very small.

Skimming code is a legitimate task. At a large/FAANG company you will often have a 400k sloc code base and only a vague idea which bits of code you need to study more intensively to find out what you need.

> The more I work with parsing, parser combinators and writing grammars for little languages, the less often I find myself using or wanting to use any regex at all.

Surprise: The most common parser combinator libraries do backtracking. That's exactly the problem. Any solution as widely used (if not overused) as regular expressions ends up exposing a number of dark corners where the design isn't as clean and tight as you would want it. There are lots of better ways, but most of them are specialized and are totally unsuited for significant areas where people need something.

That said: yes I've used LR(1) parsing (not LALR) using a library that uses parser combinators with a good interface, and it's more powerful than regex and worth it for the right usecase.

> to turn one of those kids into someone worthwhile

The basic insight here is: kids with learning disabilities, or who have a home environment not conducive to learning, are worthwhile people. And school as the institution taking care of those kids should do more than just send them to another building to rot.

It's a very straightforward thing to want your upper middle class privileged kid get all the support they need to become a future surgeon, but school is just very bad at recognizing and enhancing the chances of people who aren't neurotypical white privileged background kids.

Enforcing standardized testing as a key KPI isn't really the right solution here, but there's enough examples outside of the US how this can be done better than just writing off people as "not worthwhile".

So, delegating both responsibility and power to people with an agenda (because those are the only ones to step up when they get formal power, in addition to or rather than just informal responsibility), rather than adopting practices of de-escalation that everyone is responsible for? Paring down non-constructive and non-called for should be the responsibility of everyone in the room and not a special distinct set of people.

I'm all for training project members (i.e. people who are involved in organizing conferences or summer schools) in practical techniques for better organizing the community, and I think holding everybody responsible leads to better results than having activists with an agenda write the rules and just delegating authority to them.

I was assuming someone had an attack on the Alpaca altcoin.

So if you throw in a random word, it could be either or all of a garage rock band, an indie game, an altcoin, or a software vulnerability. And probably there are three or four academic software systems with that name...

Most likely they wouldn't stay at home but either would start their own small business (which would be more competitive now since people with employees have to pay them a livable wage) or, if they can be supported by their partner and not work, do something that's important for society but doesn't pay well (e.g. caring for your family, or for people in your community, writing, creating art).

I'd call it a wrong conclusion to assume that the thing that people do to unwind after a busy workday or what people do who are unable to work due to depression is the same as what people undertake when they can sustain themselves without spending the majority of their waking hours commuting and working.

Agree that it's exciting to see, but what's the actual potential? Flutter developing into a latter-day Macromedia Flash?

Ok, I've lived in Schwaben but only ever visited Bavaria, and never ate Pretzels in the US. So Bavarian Brezeln are not completely hard and dry but they're definitely less squishy and soft than Swabian ones.

If you have a Masterclass subscription, they have a course on making sourdough bread (from the owner of a Parisian bread shop that has been making sourdough bread for some >100 years).

Or Joshua Weissman has you covered https://www.youtube.com/watch?v=jJpIzr2sCDE

One subtle difference/fact: Brezeln are different between Bavaria (southeast) and Schwaben (southwest) - soft Brezeln are common in Schwaben, whereas Bavaria mostly has the salty hard stuff that was the parent of the salty hard stuff you can buy as a specialty in the US.

there's no operator precedence if you don't have (multiple) operators that could precede each other. In LISP-like languages these are simply functions (or more correctly, forms) which have other expressions as arguments, like any other functions or forms. LISP works just fine without much of the things we take for granted in ALGOL-like languages.

A wallet is two things:

- a private key or set of private keys which are used to sign transactions - hardware wallets store this inside the "hardware wallet" thing

- a program that interacts with the Bitcoin/Ethereum network, can display balances on a (public) key or make transactions that need to be signed by a private key (those are different, you'd use e.g. Electrum for Bitcoin and Metamask for Ethereum)

Obviously both of these can be the same (you can store your private key in a file accessed by Electrum), but to reduce the probability of getting hackes people often separate the private keys from their wallet programs.

A coin-denominated account on an exchange is often also called a wallet, but it's just an account on the exchange like you'd have one in your bank.

err ... it's kind of what you'd get from Textkernel or LinkedIn Recruiter, probably at a similar price. But with nicer fonts, so ok.

does gold qualify as a ponzi scheme? Tesla stock? Google stock? 2000s dotcom stocks?

There is a difference between overbought assets and scams, and it usually goes along the line of, the easier it is for a small group of people to walk away with all the winnings the more likely we are to call it a scam.

Cryptoassets are a suitable vehicle for scams just as bits of paper with holograms on it or devices with touch screens can be - all it takes is a suspension of disbelief and caution which we also see with overbought assets. And there's a fair share of altcoins and ICOs that look quite scammy to educated people, while BTC/ETH/LTC and some others have survived a decade of people trying to poke holes into them.

Google did an auction: https://www.cnbc.com/2014/08/19/es-took-off-but-the-auction-...

For many pricing problems, auctions are the answer. Doing a roadshow and tickling investor's interest is a separate problem, but many startups already have a size where they are well-known to a big enough share of investors.

It highly depends on the corpus and you can check for yourself.

I'm being lazy, but the results for Gutenberg books you can check online at http://labs.statsbiblioteket.dk/dsc/

- man is to woman as doctor is to reprovingly (nurse is the first noun, on position 4) - woman is to man as doctor is to snodgrass (after a couple nonsense/rare words)

The most important thing that teaches us is that big corpora (bigger than PG) are essential for this method.

How about: US citizens saved $34B by not having to stay in the highest COL areas?

Why should we sympathize with a minority of rent-seekers?

There's a difference between selling flowers in a shop and randomly going around and pointing your gun at people holding flowers and asking them to pay up. The alternative to buying flowers is just spending your money on other things, which is different to the alternatives you have in an armed robbery.

Patent law gives patent owners a monopoly (i.e. the gun) over how an invention is used, but just like there's a difference between self-defense with a gun and using it for armed robbery, there should be - and increasingly is - regulation that forces patent owners to act as market participants offering something valuable rather than street thugs going for protection money.

It's fine to dislike patents, just as it's fine to dislike anti-competitive or outright dishonest practices around patents. Either alone or both in combination are reasonable standpoints.

> and they built it all on the back of open source.

That is exactly the point of permissive open source licenses. Freelancers and commercial companies can build stuff on top of it that makes them money and brings real benefits to paying users.

Windows, MacOS, and iOS/Android would all be stinking piles of hot garbage if it weren't for open source, but then again Linux or the *BSD operating systems wouldn't be where they are today without contributions from companies acting to make a profit.

In late 90s/early 2000s the mainstream thought around numerical optimization was that it was easy-ish when it was a linear problem, and if you had to rely on nonlinear optimization you were basically lost. People did EM (an earlier subgenre of what is now called Bayesian learning) but knew that it was sensitive to initialization and that they probably didn't hit a good enough maximum. Late 90s neural networks were basically a parlor trick - you could make it do little tricks but almost everything we have now including lots of compute, good initialization, regularization techniques, and pretraining, was absent in the late 90s.

Then in the mid and later 2000s the mainstream method was convex optimization and you had a proof that there was one global optimum and a wide range of optimization methods were guaranteed to reach it from most initialization points. Simultaneously, the theory underlying SVMs and CRFs was developed - that you could actually do a large variety of things and still use these easy, dependable optimization techniques. And people hammered home the need for regularization techniques.

In the late 2000s to early 2010s, several things again came together - one being the discovery of DropOut as a regularization technique - and the understanding that it was one, the other being the development of good initializers that made it possible to use deeper networks. Add to that improved compute power - including the development of CUDA which started out as a way to speed up texture computation but then led to general purpose GPU computing as we know it today. All this enabled a rediscovery of NN learning which could take off where linear learning methods (SVMs, CRFs) had plateaued before. And often you had a DNN that did what the linear classifier before did but could learn features in addition to that - and could be seen as finding a solution that was strictly better.

But the lack of global optimum means that - even with good initializers and regularization packaged into the NN modules we use in modern DNN software implementations - the whole thing is way more finicky than CRFs ever were. (It would be wrong to say that CRFs are trivial to implement or never finicky at all, just as many well-understood NN architectures have a good out-of-the-box experience with TF/PyTorch etc. - so take this as a general statement that may not hold for all cases).

What pain points would it have saved to write the scientific paper in org-mode? Usually there's a dozen dirty little corners where you want fine-grained control and the org-mode-to-latex converter isn't giving you that by default.

I once wrote a short report that I (probably) could also have written in (any equivalent of) Word or LyX/TeXmacs, or some random WISYWIG editor. Very little finetuning of the layout required or necessary, people still liked it better than plaintext (it had some headlines and tables and stuff, but nothing too fancy). So, nice output yes but be careful if what you need is actual LaTeX document preparation.

porn already works within a well-defined legal framework. There are model releases, people getting paid etc. So if a reputable porn company interacts with their models and their consumers, everything is well. However, there are still non-legal reasons that make it a special case - see the ancestor post about a bank wanting to dictate what a merchant can offer to their credit card customers, or the post about models being treated badly (not getting paid or having to do things that they didn't sign up for). In that case, there still seems to be a difference to non-porn modeling and acting in terms of standards and of recourse available. part of this is that there are more people paying for badly made porn than for badly made action movies