blip54321

A professor in college had two terms: microprogramming and macroprogramming.

Microprogramming was the kind of thing where you're trying to, for example, cut down an assembly loop to run in 6 cycles rather than 8, design a numerical algorithm, an operating system scheduler, or similar. This is very mathematical in style of thinking.

Macroprogramming is what happens when you write a database-backed web application, for example. It's much more linguistic. It's about understanding other people's code, writing your code to be understood, and gluing a lot of stuff together.

Both have their place. Macroprogramming is taking over now, but I enjoy microprogramming more.

I hate installing and uninstalling apps. And overly permissive apps are a good sign they're not my friend in the first place.

The term for this is transfer learning.

https://learningcurrents.weebly.com/visible-learning-surface...

It takes a minimum of about 3 semesters to be able to get linear algebra to the stage of transfer learning. There are slower ways to do it (which most people follow), but there aren't shortcuts -- there is no way high school Algebra 2 can get you there in the time allotted.

That kind of high school surface learning is not a step you can skip, though:

"Surface learning is critical because it provides a foundation on which to build as students are asked to think more deeply."

There are a lot of shortcuts which can cut back on the number of hours, mostly by spreading out the number of years. Early surface exposure helps a lot, as does revisiting topics in different contexts.

https://en.wikipedia.org/wiki/Spacing_effect https://en.wikipedia.org/wiki/Spiral_approach

Since finishing school, I've used very little of the math I learned in school.

I've used a lot of math. Most of it built on high school algebra and geometry. Does that mean that's all I needed to learn?

No.

The key thing I do use is more ephemeral: Mathematical maturity. In my current job, I use math I never learned in graduate school. I'm able to learn it quickly because I learned a lot of math back then. My math classes were a way to develop mathematical maturity.

Which specific math I learned in graduate school was almost incidental. What I picked up was the ability to learn new math.

After 9/11, a clever MIT undergrad grabbed some form of alqaeda.net. Any email sent to the address went to the corresponding @mit.edu email address. You could email professor_alice@alqaeda.net, and it'd arrive at professor_alice@mit.edu.

Undergrads sent emails like that for the lols. Recipients got freaked out they'd end up on some government watch list.

I think this would be a better state than we have now...

... but I feel like it isn't the ideal solution. I don't think we should make the perfect the enemy of the good, but what I'd really like to see is that people own their data about them.

I should be able to figure out what vendors have data about me. I should be allowed to be forgotten. There should be severe liability if they don't forget me.

I should be able to contract away my data, but in return for a real benefit, and that shouldn't extend to sale / redistribution.

I don't buy anonymization. De-anonimization is too easy, too good, and progressing too quickly.

Indeed. It's an arms race. Search engines versus SEO.

This isn't uncommon. For example, when MIT/Harvard took over edX from the original researcher who built it, they found they didn't know how to build software. The new team introduced this same issue: passwords in log files.

It was fixed much later. You can look over the git logs.

I'm not sure this is so much of a "how good this happen," but a "thank you for being transparent." Most organizations cover this sort of thing up, like MIT/Harvard (and now 2U).

Good of github to announce this openly!

Reasonable people cannot disagree about the framing of GDPR as a human rights law. The second sentence is "This Regulation protects fundamental rights and freedoms of natural persons and in particular their right to the protection of personal data."

Reasonable people can disagree about the extent to which privacy is a fundamental, human right, or where the bounds are, but that is literally the phrasing of the law.

Reasonable people can argue about a lot of issues, and views on rights change with time. Ancient Greeks and not-so-ancient Afghans had sex with kids. Just over a century ago, women couldn't vote. It's hard to predict how views on human rights will evolve. Right now, there are huge cultural disconnect about a lot of things digital. It's not clear where they'll land.

Jurisdiction issues are complex. In this case, the jurisdiction is defined by the location of the customer, not the business.

If your business ignores EU courts, that might not have an immediate impact, but in the longer-term, you have a liability if you ever do business in Europe, want to be acquired by someone with a business presence in Europe, and potentially in the future, travel to Europe.

GDPR is framed as a human rights law, and that has long-reaching claws.

It is currently not well-enforced, but there are many examples of clawbacks coming in. For US slavery, those clawbacks are coming 160 years later: buildings, businesses, and schools are being renamed. Statues are being torn down. In some cases, you're starting to see reparations (see Harvard). Milder versions of racism are subject to cancellations; things acceptable in 1980 are having repercussions on people's careers in 2020.

Then you've got issues of when you're persecuted for an unrelated reason, and the government is looking for an excuse or pretext to take you down. A famous mobster was taken down a century ago for tax evasion.

V3 is, unfortunately, nowhere near usable.

It'd be as if Microsoft had discontinued the Windows 95 user interface in favor of Microsoft Bob, in 1995. It's that not-done.

I'd be opposed to the transition if it were done, but V3 is nowhere near workable. Documentation, APIs, the whole shebang. It's just unfinished.

I feel like the materials are whatever is found on site.

That may mean that a machine may need to be able to sort, smelt, sinter, form, ram, engrave, or otherwise process whatever it found on-site.

That's definitely non-trivial, but in the sense of making a CPU with millions of transistors, and not in the sense of Mt. Rushmore. The R&D costs will be astronomical, but there isn't much there which is /fundamentally/ expensive, at least so long as you're not going to deep.

I can describe a number of beachheads and early markets -- places where you don't need all of that at one, or ultralow costs. I think the Boring Company has a decent one.

I'm really not sold on the costs of digging down being all that high, long-term.

I'm 100% sold on it being expensive today, but I can come up with ways to drive them /way/ down with automation. The critical thing is:

1) You don't need materials.

2) You don't need to transport anything other than a digger

I agree it's hard, like engineering a CPU was hard, but I think Elon's got the right idea with the Boring Company. It's not /fundamentally/ expensive. Fundamentally, building in-place with available materials should, some day, be cheap.

CFAA has actually narrowed quite a lot.

The law isn't crazy as written; it's crazy as misinterpreted. Van Buren v. United States is the Supreme Court case that tossed out most of the batshit crazy interpretations. Today, MIT wouldn't be able to Schwartz people as easily as it did in years past.

(And yes, MIT kept threatening folks like whistleblowers under the CFAA to get them to sign NDAs long after Schwartz; to Schwartz someone was a verb around MIT)

The intent is very much territorial. I don't think your interpretation is 100% correct, but I think it's 95% correct. Jurisdiction is clearly defined as territorial:

https://gdpr-info.eu/art-3-gdpr/

However, the framing as a human rights law, protecting "fundamental rights and freedoms of natural persons," means that the intent is somewhat more broad:

https://gdpr-info.eu/art-1-gdpr/

It defines principles which are believed to apply everywhere.

That's reinforced by language like "This Regulation applies to the processing of personal data by a controller not established in the Union, but in a place where Member State law applies by virtue of public international law." This extends jurisdiction as far as practical, in ways which are legally ambiguous.

"Legally ambiguous" generally means that you might be right, but the cost of finding that out will be astronomical.

Human rights laws have an inherent friction to them. On one hand, if random dictator's private business violates human rights laws, that's outside of jurisdiction, so not much can be done about it. Sovereignty is an important principal. On the other hand, it's clearly viewed as not okay, and often has some ramifications at some point. If you're applying for a contract, and have a track record of human rights violations....

I wouldn't quite go so far, but I don't think piecewise enterprise SaaS makes sense. The whole model of having dozens of interacting SaaS pieces, any of which make change APIs, have a bug, or go under any day has peaked.

I'm okay relying on AWS.

I'm not okay having one company handle email, another columnar database, another tabular database, another map-reduce, another streaming / logging, and so on. I understand why each of those might be better than their AWS counterpart, but for business continuity or for security, it's a train wreck.

I'm usually okay relying on hosted versions of open source platforms, since if something goes very bad, I can move over to hosting them myself. Even there, I hesitate with things really central to continuing operations. But for the 95% of other stuff, I find something like hosted postgresql or redis to be most robust.

A company specializing in redis will beat my IT staff. On the other hand, it's low-risk. A lot of companies disappear in a few weeks, but it's rare that they disappear overnight.

Proof is not required. In the US, for example, you file taxes on an honor system.

If you've lied, there are periodic audits, and you might get thrown in prison.

I assume you can lie going to China. If you do and get caught, you'll probably never be welcome to visit China again (or fines, or prison, or some other consequence; but most countries would just keep you out).

Personally, I feel like there are too many dollars and startups in this space now.

I feel like the risk profile is such that I'd never buy most of the products being built. I don't want core business infrastructure sitting in the cloud of a startup which might not exist tomorrow due to missing a VC round or an agile pivot. I also don't want an integration of 50 different cloud services.

I'm glad to use established open-source technologies. I'll also use AWS or similar big players if open source doesn't exist. However, most of the niche proprietary startups just don't make much sense here to me. I'd invest a lot of money and take on a lot of risk.

I do feel like there is big money in value-adds: hosting open-source solutions, consulting, etc. A lot of non-tech companies are struggling with data, ML, and visualizations.

There's a feedback loop here. Once these startups start shutting down, the above problem will be recognized, cascading their collapse.

This is incorrect.

There are complex laws governing residency which vary country-by-country. It's not atypical that, for example, if you spend 183 days in a year somewhere else, you're no longer a resident.

You're generally only a resident in one country. This governs taxation -- taxes are primarily based on residency, with some caveats. As a US citizen living in the EU, you will not pay the same US taxes as you would if you were living in the US, and in many cases, zero taxes.

GDPR is very clear -- EU citizens living outside the EU are NOT covered. US citizens living in EU are covered. That's not uncommon for laws; I don't get a ticket for breaking US traffic laws in the EU, or vice-versa. Very few laws reach across borders like that (taxes, some forms of child abuse, some forms of bribery, child support, arms trafficking, some types of military service, espionage, etc.).

However, as the other poster pointed out, from the perspective of a tech vendor, understanding the difference between:

- EU resident

- EU resident using a VPN in the US

- EU resident traveling in the US

- EU non-resident living in the US

For each request which comes in is practically intractable.

GDPR is also framed as a basic, universal human rights law. That also can have unintended tentacles. If you don't want a liability hole, it makes sense to honor GDPR for everyone, in practice (even if not under your ToS).

Practically, that's what everyone does. I've never had a GDPR request declined on the basis of residency (the last qualification being important).

GDPR applies to residents of Europe.

That said, I've never had a problem using it from the US.

I think that's a good starting point. Any good references? I'm looking up the one already given in thread.

I have a lot of experience with medical biohazards, high voltages (>2kV), welding, dangerous tools, handling dangerous liquids (HF and similar), etc. I've never hurt myself. That's enough background to know not to do things before I know how to do them safely.

I'd like to mess with them at home, but not at "the necessary volumes" to go through a foot of steel as in the video.

You know those little bang snaps you throw at the ground and they make a popping sound? For DIY, ideally, I'd do things in that kind of volume. I'd probably need a smidgen more, but I wouldn't want to do more than e.g. make a pinhole in 29 gauge sheet metal.

What I'd really love to make -- but I don't think I ever could safely -- is a little 1/8A or 1/4A rocket engine (of the that cost a couple bucks at the local hobby store).

This is science lab with kids, not bomb-making 101.

Question: Is there a way to play with these things at home? Any pointers to DIY stuff?

I think the line is when your troops shoot their troops.

Providing weapons has been a US-Soviet past time for the whole Cold War.

In this case, I think the US could have even moved in troops without being at war, early in the invasion. It just needs a pretext. By far the best response a week in, when Russia claimed chemical weapons, Nazis, etc., would have been to send in a multilateral force to investigate Russia's claims. It's face-saving for the Russians and ends the war.

Tensions are too high for that now.

For the most part, because you'd be IO-bound, and you'd be limited by single-threaded speed.

A better architecture is to have the same, but as a co-processor for highly multithreaded tasks. That's basically a GPU. You get a few thousand cores on an NVidia 3060, in an architecture designed for this.

To the other comments in the thread:

- I'm mostly giving big-O / ballpark numbers.

- In terms of power, Pentium IV was 180 nm - 65 nm. A 7 nm process can be much more power-efficient.

The trick isn't to shop for /most/ Chromebooks. The trick is to shop for /decent, expired/ Chromebooks. Chromebooks are designed around planned obsolescence, and all come with a use-by date, after which they stop updating:

https://support.google.com/chrome/a/answer/6220366?hl=en

Chromebooks near or past the planned obsolescence date can be had for a song, including decent models. The market is close to non-existent, so there's a glut of them.

My Chromebook has a 3200x1800 display, 16GB RAM, and takes an SD card (for expandable, albeit slow, storage). That's plenty for most of the types of work I'd like to do on a boat. It was under $200, almost expired. New, it would have been close to a grand.

The most popular way to install Ubuntu is with crouton:

https://ubuntu.com/tutorials/install-ubuntu-on-chromebook#1-...

However, I installed it natively. Here's a random tutorial:

https://dbtechreviews.com/2018/09/how-to-install-ubuntu-on-c...

The key annoyance (really the only difference from a "real" laptop) is you have to hit a special key sequence on every boot.

I definitely don't think of it as a "world of pain." I wouldn't use it as my primary laptop, but it's great as a device I can use in places I'd never take my primary laptop.

Expired Chromebooks are cheap, and great for installing Linux.

For my purposes, a $100 used Chromebook is perfectly adequate, and is the sort of device I can take on a hike, kayak, or bike ride, and not worry if it's lost, stolen, or damaged.

Cores are basically free.

Interconnect is expensive.

Seriously. A Pentium IV was 40M transistors. A Ryzen V 2000 has around 5 billion transistors. It could fit 100 Pentium IV cores if desired.

That's not desired -- those transistors are better spent bumping up IPC a little bit -- but we can have a perfectly adequate processor at 1% of a modern CPU.

Pentium IV single-core performance is almost identical to a modern entry-level netbook processor:

https://cpu.userbenchmark.com/Compare/Intel-Pentium-4-300GHz...

Thank you!