solresol

It's pretty rare for a child growing up in an anti-science household to be convinced to change their worldview by abstract philosophy, not least because they won't have access to those sorts of books.

It's quite common for a child growing up in an anti-science household to be convinced to change their views on science by devout Christians writing about evolution, dinosaurs, astronomy and so on. Fundamentalist parents are often quite surprisingly happy for their children to read those kinds of books because they are written from the perspective of the in-group. There's a reason that Life of Fred is popular.

If you want to write a book about the theory of science and what truth is, go ahead, but you aren't going to make much of an impact on that particular target audience.

I'm working on it, but with a loss function that penalises hypernyms/hyponyms less than other kinds of mistakes. I'm vaguely optimistic that this is going to be more efficient.

I've got video recordings of my lectures. I haven't edited them properly yet, but I can share them with you privately. Email me (it's in my profile) and I'll get them to you.

If you care about inference time, then you'll do two things.

1. Train a student model from your fine-tuned model. (Known as "knowledge distillation").

2. Quantize the student model so that it uses integers.

You might also prune the model to get rid of some close-to-zero weights.

This will get you a smaller model that can probably run OK on a CPU, but will also be much more efficient on GPU.

Next: architect your code so that the inference step sits behind a queue. You do not generally want to have the user interface waiting on a inference event because you can't guarantee latency or resource availability, and your model's inference processing will be the biggest slowest thing in your stack, so you can't afford to overprovision.

So have a queue of "things to infer", and having your inference process run in the background chomping through the backlog, storing the results in your database. When it infers something, somehow notify your front-end clients that it's ready in the database for them to retrieve. In this model, you can potentially run your model somewhere cheaper than AWS (e.g. a cheaper provider, a machine under your desk).

Or, for the genius move: compile the model to ONNX and run it in a background thread in the users' browser, and then you don't have to worry about provisioning; users will wonder why their computer runs so slowly though.

The expired domain abuse thing --- that's interesting because finding that signal was one of the few things I got done while I was at Google... in November 2006! And I vaguely remember being told that the fix went into the codebase after I identified that it was happening.

General advice, true for most people. But for some people, the only way they reliably lose weight is through increased exercise.

Likewise, I never expected that there would be a car-free village in Hong Kong. It's peaceful and quiet and laid-back: exactly what you wouldn't expect in Hong Kong.

You can also get midi interfaces for cathedral organs; letting you use the organ console to control midi devices, and also letting midi devices control the pipe relays in the pipe organ.

Of course, whether you would want to use the OP's app in some giant cathedral is another question.

Like a Boox Palma? That doesn't have the SIM ripped out, it's a phone-sized device that isn't a phone.

It has wifi though, but you can't watch videos or scroll fast because of the e-ink screen.

Finish your degree. You won't regret it. At the very least, it will make migration to another country a lot easier.

The professor who failed you may have never seen SQL Server before teaching the course. If they are an adjunct, they have very little control over what is offered to them, and many will take teaching jobs without knowing much because there are no consequences (to them) for doing silly things like you saw.

Choose courses with professors that have something useful to say. Look for older professors that worked in industry first and then came back to academia recently; then look for professors whose lectures other students have enjoyed.

Your degree may have some "core units" that you have to take, where you will encounter incompetent professors. The secret is: no, you don't have to. You can get just about anything waived. If you talk to the course convenor and map out a curriculum for yourself that is filled with weird random topics, and you can say why this is the right sequence for you, they will almost definitely say yes.

Look really far afield, and find something else at the university that interests you: philosophy? Esperanto? Demography? Paleobiology? You have the advantage that with your tech skills, you don't need the degree to gain a foothold into industry and get a job: you need the degree to demonstrate that you can complete things, are smart, can learn new things, and are willing to signal delayed rewards. So it doesn't matter what the degree is in for you. There are an endless number of interesting topics out there. Find something interesting and run with it.

What you describe does not match my experience at all.

Occasionally I think the medical world has a communication problem.

In 1989 I had some blood work done which checked for Epstein-Barr (it wasn't present) for reasons that made sense at the time. Then I got ME/CFS a few weeks later, and blood work showed that it was present this time.

That's not a slam-dunk for Epstein-Barr as a cause, of course, but I thought it was interesting enough to have the data that I talked to my immunologist about cowriting a paper on it.

She demurred, saying that quite a few such papers (where patients by chance had had before-and-after tests) had been published and there wasn't any value in adding another one, since it was pretty common to see the same pattern of Epstein-Barr virus not being there beforehand, and then appearing after in blood work after the onset of symptoms.

Fast forward to 2023 and now there's a paper saying that ME/CFS might have a post-viral infection origin.

Both the services companies I started were profitable within 8 weeks. (6 and 8 respectively.) One was sold 14 years later, the other is still going.

Services companies are very fast to get started, and very easy to make profitable.

Building a product company on the other hand...

The article didn't mention UTF-9 or UTF-18 for when you want to do unicode on your PDP-6 or Univac. https://datatracker.ietf.org/doc/html/rfc4042

A reviewer for a scientific paper I wrote asked if I had considered the impact of UTF-9 on what I was doing. To this day I don't know if they were trolling me.

You should have seen the platform it replaced. ServiceNow was essentially "let's rewrite HP Service Manager from scratch to remove the legacy debt." Service Manager was what you get when you develop a mainframe green screen application using a low-code RAD system and then try to maintain it for 20 years.

Amusingly, I'm consulting with a company now whose business model and product strategy is "a rewrite of Service Manager that's cheaper and more sane." Presumably the cycle of rewriting these kinds of platforms will continue until the heat death of the universe.

Mostly churches. A few universities where there was a conservatorium of music had some smaller organs.

The company I worked with didn't install into anyone's home, but one of our competitors did once.

When the musician presses a key on the console, there are three possible technologies for turning that key press into air going into the pipe.

1. It could be an electric signal, in which case the most common thing to fail in the organ is the relay that opens and closes the valve.

2. It could be a pneumatic signal: the keypress opens a valve to let air through to a tiny tube which opens the valve on the music pipe. This was surprisingly reliable, but I'm slightly biased because a few decades before my apprenticeship, a slightly idiosyncratic pipe organ builder had been operating in the area, and he was a passionate believer in the goodness of pneumatics over all other technologies, so over his career he got very, very good at making reliable pneumatics. On those organs, it tended to be the tubing that would split.

3. Fully mechanical: you press the key to open the valve to let air into the pipe. The most stops you have open, the harder you have to press the keys. These tended to be much older instruments, so I saw a lot of holes in the bellows material.

---

Other surprisingly common failure: birds and mice falling into the big pipes. Then they can't get out, and die in there. It gave a strange muted sound to that one note.

Nobody knows. It started in about 900AD in monasteries (because the monks were the only people technical enough to maintain and play one) and became part of church tradition a few centuries later.

And after a millenium of use, in the late 1980s to early 1990s almost all churches stopped using them.

You take out the pipes, fill with them with packaging material, and move them out of the way. Then you unscrew the racking for it, and disconnect all the airflow tubes. Put that somewhere. Keep going until there's nothing left.

Then you either get a very large truck to carry the console, or you pull it apart too.

Generally you make the apprentice do this. I was the apprentice. I got very good at carrying 64-foot pipes from one end of a church building to another.

Every significant pipe organ is unique, and custom-built. You can modify and change it: e.g. add an extra rank, move the pipes around to make it fill the room better and so on, so even if two organs started the same, pretty soon they are different.

I did a handful of moves. It made sense when a church was moving to a new building, or some other "nearby" relationship, but for the most part it wasn't that much cheaper than buying one new. By the time you cushioned the pipes for transit and transported them somewhere else, you could have received an order from a factory for new pipes. The little tiny pipes are almost guaranteed to break or be deformed on a long move, and all the leather around the bellows will get torn.

The organ in the parent article is big. That's going to be a large job just moving it, and there aren't that many buyers for organs that big. (Because it will be expensive to maintain.) So they are looking for a major cathedral, a university's great hall, a big city town hall or something like that. All of which are going to struggle just to fund the work assembling the organ, let alone paying to buy it.

There was a requirement that the organ be kept together, which is why it is valueless. If they didn't mind someone buying it to fill up a spare parts supply, they could probably get some money for it.

Feel free to AMA about this (organ building in general; I have nothing to do with the site specifically). Before I moved into tech I did a pipe organ building apprenticeship.

Maybe I was just early in the game, but yes, I've got access. I didn't have any particular project that I wanted it for.

Email address in my profile.

I have heard this version of events many times, but that's not how I remember it.

I remember the only international flights allowed as being the repatriation flights. (e.g. when an overseas government organised for some non-Chinese who were in Wuhan to be able to come home). And presumably the return flights for planes that would otherwise have been stranded in Wuhan.

FactCheck matches my memory...

https://www.factcheck.org/2020/05/trumps-flawed-china-travel...

Wikipedia for 23rd January says all outbound flights were banned:

https://en.wikipedia.org/wiki/Timeline_of_the_COVID-19_pande...

And anyway, there's something wrong with the whole idea: if domestic flights were banned international flights were still allowed, wouldn't the people of Wuhan who wanted to get to Beijing just fly via Bangkok or Seoul? A ban on domestic-only flights wouldn't have achieved anything.

New Scientist (but I read the print version).

Terraform, Helm (or just plain Kubernetes config), Ansible, other devops technologies.

This would be the most beautiful experiment.

Somehow we would need to track to two code bases for a decade afterwards and see which was the better choice.

I had various bowel problems while on it, which were slightly unusual and hard to relate back to taking metformin, but they all cleared up when I stopped taking it.

It's a very well-known drug, and a lot of patients have taken it for a long time. Side-effects are rare.

Niches = riches.

You want a super-tight thing that you do (e.g. "I fix performance issues for large NodeJS code bases", or better still "I write tooling to measure performance hotspots in NodeJS code bases".)

Despite this sounding like cutting off potential customers, it doesn't work that way. Customers will reach out to you with some generic work that they want done "because I know you know NodeJS" even when you say you specialise in something.

A good first book to read: "Book yourself Solid" by Michael Port.

He's a specialist in bankruptcy and liquidation administration. Essentially the debtors have appointed him to take over and identify what can be sold or salvaged. He gets paid out of whatever funds he can find.

> Prompt: lecture notes about p-adic machine learning

In p-adic machine learning, the p-adic numbers are used to represent features. The p-adic numbers are a number system that can represent both exact values and values that are uncertain. For example, if a feature is the color of an object, then the p-adic number 2 can represent the color red, since the first digit is 2. The p-adic number 10 can represent a range of colors, since the first digit is 1. The p-adic number 20 can represent a narrower range of colors, since the first two digits are

So confident and yet so completely wrong.