retrac

Of course it's not ideal. But you're really getting waylaid.

Here in Canada (where we supposedly like restorative justice and giving people second chances) the penalty for murder is life in prison.

The non-parole period (the amount that must be served) is set by the judge at sentencing and ranges from a minimum of 10 years for second-degree murder. (The longest period is never: indefinite incarceration until proven to be safe around others -- mostly reserved for serial killers.)

That is to say: the absolute shortest sentence you can get in Canada for murder is ten years.

In America I believe it's often many decades if not the death penalty, no?

Even Norway - renowned for the progressive humanity of its legal system - mandates at least 8 years jail for murder.

In just about any country anywhere 10 years for murder is, objectively, a relatively light sentence for murder.

Not at all sure what the sentencing for murder has to do with firearms, really. Other than firearms being an effective way to murder. Most murders in my country are beatings and stabbings, probably since firearms are not as widely available. Does that prevent a few? Probably. Still seems non sequitur.

EDSAC as early as 1949 had an "initial orders" module. A boot-up ROM. There was a bank of rotary selector switches, to set the octal digits in each ROM word.

David Wheeler (of subroutine fame) figured out the most sensible thing to put in the tiny ROM was a paper tape loader and mini-assembler that loaded the rest of itself from paper tape. So at switch-on the machine would start reading the tape input and run. Programs could be written on the teletype with mnemonics and relative addresses in octal.

It seems a little strange that more advanced later machines like the original PDP-11 from 1970, couldn't "autostart" from ROM like that. You had to toggle in a bootloader.

I think that was because core memory was non-volatile. (EDSAC didn't use core; it had a refreshing DRAM-like delay line memory.) With core, you only had to toggle the bootloader into once and, provided your software didn't accidentally overwrite it, it was still there after a power cycle. So re-toggling in the boot loader was not really that common, despite how much lore surrounds it.

> Consider little tricks like looking at urls and trying others that fit the pattern

https://www.cbc.ca/news/canada/nova-scotia/teen-accused-foi-...

https://www.theregister.com/security/2018/05/07/hacking-char...

https://globalnews.ca/news/7590375/ns-foipop-website-back-on...

> But there have been no wars between major powers in the last 80 years. Maybe people in power have realized that all-out wars are no longer productive.

People talked like that before WW I. There hasn't been a major war in Europe since 1870 almost half a century and trade is such a large war would bankrupt all the major powers and so would never happen. The Great Illusion was published in 1909 and was positively received at the time:

> In The Great Illusion, Angell's primary thesis was, in the words of historian James Joll, that "the economic cost of war was so great that no one could possibly hope to gain by starting a war the consequences of which would be so disastrous."[4][5] For that reason, a general European war was very unlikely to start, and if it did, it would not last long.[6] He argued that war was economically and socially irrational[7] and that war between industrial countries was futile because conquest did not pay.

It seems to be a vibe coded joke that implements a database in the worst way possible. For example, it appears to store post data in the timestamps and permission bits of the file system.

"Agent" was already established lingo in computer science and AI when The Matrix was released.

https://dl.acm.org/doi/10.5555/648203.749270

Practically? Quite a few codebases from *nix systems rely on the case sensitivity. To do a successful git clone you need a file system that handles both Makefile and makefile, or whatever.

I ran into it more than once back in the day when I used Mac OS X as my primary OS.

More generally -- case sensitivity is a conceptual nightmare in the Unicode era. Should Cyrillic or Greek be case-insensitive as well? Etc. Do you really want the full complexity of Unicode string handling in your file system?

I would suggest treating file names like raw bytes. On modern Linux, anything but NUL is valid.

> Is this about the word "understand"? We're past that discussion ..

We're really not.

https://buttondown.com/maiht3k/archive/how-to-talk-about-ai-...

Very cool. I've been working on something similar. My cutoff date is 1994 as Usenet changes substantially in nature then.

May I ask how you collected the data? Usenet is not fully archived anywhere. No one has a complete set.

I am aware of:

* the NewNews disc series on CD from '91 - '93 which has a substantial text feed for the big groups during that era (most of these are on IA.org, but some are not!)

* The Dejanews archives via Google up to c. 2013. Also up on IA.org These are lacking large amounts of early posts, and are highly spotty before about 1994 (when Dejanews started).

* The Utzoo tapes over at https://archive.org/details/utzoo-wiseman-usenet-archive cover a part of '81 - '91 but only from one host. "Far parts" of Usenet over in Japan or Europe are barely covered.

* Some groups preserve per-newsgroup histories scattered all over and tending to drop offline over the years

* fj.* is well-archived but it's Japanese

The chart speaks for itself I think as to coverage: https://i.imgur.com/WCQm4od.png

I got so desperate as to go scanning zip and tar files on old ftp sites for anything with a Usenet header. That scored a few hundred thousand articles mostly from the 90s.

The big spike in the early 90s is from those CDs archives in the months where I had access. It gives an idea for how limited those Giganews/Google and other surviving archives are.

I figure less than 50% early text Usenet survives, unless someone has vast hordes of data that they are not disclosing.

As for the Utzoo archives - they're legally encumbered and some of the posters sue anyone who hosts them. (They're not available from the IA - taken down.)

You have a takedown page but the legal and copyright issues were one of the main back-burner items keeping me from making my archive live. Some of the more eccentric Usenet personalities are still alive.

Same with privacy issues. I notice you have full text search. I was reluctant to do that for privacy reasons.

I discovered very early on how to index by email but that felt quite invasive. I took that out, and built a graph of all the X-refs: which turned out to be more interesting (walking the discussion trees) and somewhat more privacy preserving in that you can explore but not search.

I was also thinking of an AI-based content warning filter, because I have a "random post" button and it is sometimes wild what that lands on.

For that type of structured data (logs and such), a custom dictionary can be extremely effective. Zstd among others support generating a custom dictionary. You just run zstd --train over the data first, and then feed that in when you run zstd. For example: I found ~10 gigabytes of Usenet headers compress to ~700 MB using Zstd and 1 MB shared dictionary -- and that's with each header individually compressed, so o(1) lookup time.

Effective natural language processing of sign languages. Current LLMs are almost completely incapable. And it's a steep hill to climb. No text corpus. Must be learned from video. I do not believe current LLM approaches are capable of this with the amount of training material available, regardless of compute. Though I'm eager to be shown wrong.

> What's In A Name?

> What is your reply when someone asks your profession? Computing Engineer? Numerical Analyst? Data Processing Specialist? To say "Computer" sounds like a machine, and "Programmer" has been confused with "Coder" in the public mind (if your particular segment of the public knows what you are talking about at all!)

> It would help our profession to be widely recognized if it had a brief, definitive, and distinctive name. This should be general enough to cover a variety of subfields--from numerical analysis to data processing, but specific enough to imply that computing applications are involved. Consider the solid professional sound of such terms as "Petroleum Engineer" or "Nuclear Physicist." What can we use that will be equally clear-cut--and at least half as impressive? So far our ideas have been supremely uninspired. Any suggestions?

> (signed) Editors of DATA-LINK (Los Angeles ACM Chapter Newsletter)

> [Several names have been suggested to represent various phases of our profession. Though the suggestions are inevitablty facetious I list them for your attention: Turingineer, Turologist, Flow-Charts-man, Applied Meta-Mathematician and Applied Epistomologist. -- Editor

-- Letters to the Editor, Communications of the ACM, Volume 1, Issue 4, April 1958

> Sextides sells research compounds intended for laboratory research use only, not for human consumption.

That seems completely incompatible with the ad copy earlier on the page where they are clearly selling these for human consumption

My friend in the United States got an ad (on Instagram) yesterday, for a service that promises to sell peptides and it includes a free monthly subscription to Grindr - a sex hookup app: https://www.woodwork.com/glp1-treatments

As we both said to each other: "This is legal?!"

Even if it isn't legal the regulators have completely forfeited all responsibility.

> I guess Cleopatra feels somewhat 'close' to us due to the surviving literary and archeological artifacts from the Roman empire.

It's due to the historiography of western civilization that for so long always traced its origins back to Greece (and Cleopatra was, of course, Greek.)

Compare to Mesopotamia, Sumer mentioned by the parent comment. There may be more surviving cuneiform literature than Greek and Latin. The sheer volume is kind of hard to overstate. The Library of Ashurbanipal is a prime find but hardly the only one of its kind: 30,000 tablets including mostly-complete copies of epic poems, histories, dictionaries, and thousands of administrative and personal letters. And there's millions more tablets from thousands of other digs.

And the crates and crates and crates of ordinary shipping receipts, if ever systematically studied, offer the potential to understand their society in ways we never will with classical antiquity.

But most haven't been read. There aren't enough people who know Sumerian and Akkadian (and all the other languages written in cuneiform over its 3000 year history) and how to read cuneiform for the task to translate them all to be completed in our lifetimes.

Stories of Mesopotamia do not always resonate. In fact they are often very alien and inscrutable. It's very hard to grasp the Mesopotamian mind. I'm unsure if this is lack of familiarity or because they really were so different, because in the western tradition they are the canonical Eastern Other, and shrouded with thousands of years of myth and misunderstanding.

Despite, or perhaps because of, the strangeness, I found reading translations of Gilgamesh and Atra-Hasis to be revelatory and hope that my descriptions don't put anyone off. They're worth it. To this day "Babylon" has all sorts of associations, many of them unsavory. Which is rather unfair to them.

The Babylonians had 360 degrees in their circles, 24 hours in their days, and when they looked up at the stars they saw patterns like the Twins (what the Greeks would borrow as Gemini), the Bull of Heaven (Taurus), and the Crawfish (Cancer), the Hebrew month of Tammuz is named after the Sumerian fertility god Dumuzid, and the Mesopotamians leave many other echoes in the later cultures, now that we know to look.

My own journey to successfully completing a squat took about three years. And I would now agree about squats.

They were always something I avoided. I have a bad back, visibly curved on the x-rays with a turned in shoulder. When I started squats I did not not have the core stability to do the exercise correctly. Not lack of strength really, but a lack of coordination and nerve adaption, and exercising all the tiny little muscles for balancing a heavy dynamic load. (Rather than stabilized loads like you get with weight machines.)

My first attempt was the empty bar which is about 35 lb -- and I almost fell over. I couldn't get my knees horizontally parallel to my waist, either. My knees wouldn't flex that much, and neither would my ankle.

So I started with unloaded exercises, using a ballet bar to balance. Fast and slow stretches. Body weight squats. Worked up to doing 100 sets of them. My knee stopped feeling weird. I then went through a phase where I would sit down on the ground and then lift myself back up with small dumbbells, from two 10 lb weights, eventually working up to the 35 lb ones.

Then I attempted the unloaded bar again. Almost fell over again. Went back the next day. After a couple months the motion was confident. Slowly started loading it. Now at 3x3 at 135 lb. Some of the younger men at the gym eyeball me like "What?" -- but the scoliosis only shows on an X-ray and I'm too old to worry about what people at the gym think. I'm very proud of it and never thought my body could do this.

I'm reminded that chips like the Intel 8080 and Motorola 6800 were not intended to be general-purpose computers, and were originally designed mostly to be embedded controllers and as building blocks for "real" computers like terminal and disk controllers.

I suppose in principle the RPi Pico could self-host its own OS and devel environment. 520 KB of RAM in the old days was plenty to fit an OS and compiler. No MMU so no paging though.

I consider most deep learning and statistical learning systems to be AI.

But I also consider traditional symbolic AI to be AI.

You only underline my original point: no one can define AI.

All of the major hearing aid manufacturers incorporate DNNs or transformers into their chipsets. The manufacturers advertise it as "AI" and that's probably a fair description.

It's a kind of filter, implemented with neural networks, sound-to-sound translation. For example Oticon's white paper on their AI based noise reduction system is here https://assets-je.cas.dgs.com/-/media/oticon/main/pdf/austra...

> > Our goal was to train the DNN on sound scenes so that it could solve the task of balancing sound sources by preserving cues and attenuating noise. The large amount of data needed for this training was recorded in different sound scenes across a wide array of listening environments representing sound scenes that listeners would typically be exposed to in their everyday lives. We used a specialised spherical microphone, capable of capturing 360 degrees of sounds to provide the DNN with a spatially accurate and detailed sound scene and to train it on the full sound scene.

They claim it improves signal to noise over traditional DSP hearing aids.

> AI use is prohibited for any use for any purpose in any exam situation.

Read literally this bans the use of hearing aids in an oral examination. All the devices on the market today use noise reduction based on neural networks or transformers. Isn't that AI? (The manufacturers' ad copy certainly insists so.) Presumably not what they intended but "AI" is not defined in the document.

> Its use of "dense jargon" and "stilted metaphor" is actually surprisingly consistent - it's speaking its own dialect, and you get used to it.

This dialect is idiosyncratic to you and Claude based on your session history and memory.

I've noticed Claude's output mimics my writing style.

> Registers the board implements but whose behaviour is not modelled

Right down to my preferred spellings.

As several comments I've read on HN suggest, this jargon which can be so precise in the mind of one person, tends to rapidly fall apart when multiple people try handling it.

It's the Bluetooth; when Bluetooth connects most hearing aids reduce the environmental volume slightly so you can hear the streaming content. Some app has started to play audio (perhaps silently). I notice it on some websites.

How much HAs reduce environmental sounds when streaming is configurable by the way, but only by the audiologist if you don't have your own programmer. I don't like the effect and prefer to just turn down the volume manually.

Ghost characters reveal something about the joint semantic and phonetic nature of most Chinese characters. I know 彁 isn't real: but it still insists on a pronunciation: ka in Japanese, (gē or gé Mandarin). And it hints at meaning: bowed weapon, sound, elder brother; but your guess is as good as mine!

> RISC-V MCUs with MMUs or MPUs

MPUs are more common. The Physical Memory Protection option is basic and allows ranges of memory to be set unavailable in user mode. A few well-defined ranges do let you lock a user process down securely but it's not a real MMU. Low-end microcontrollers don't have enough RAM to warrant a real MMU.

The RISC-V core in the Raspberry Pi RP2350 has PMP as do the ESP32 cores.

I find it fascinating how self-aware the Mespotamians were of their own literary culture and history. Later rulers would do early forms of archaeology and dig up the ruins of earlier kings. The Library of Ashurbanipal has sign lists (proto-dictionaries) which go way, way back, to the early Sumerian period, and it seems the Assyrian elite could still read inscriptions that were over two thousand years old, when Nineveh fell in 612 BC.

Bit of trivia: the Chinese word for honey 蜜 (Mandarin mì, Cantonese mat6, Japanese mitsu) may be a borrowing from Tocharian, cognate with the English word "mead".

Magnetic amps were indeed used in the V2 in the control electronics.

But they substantially predate World War II and were once widely deployed and widely understood. The technology goes back to at least the late 19th century. It was used in many of the places where we'd use an amplifier today, like to run signals in a factory where hydraulics are not practical.

They can be quite fast - if the AC supply is well into the 10s of KHz they can reproduce audio range frequencies.

Some of the first electronic radio transmitters were based on the magnetic amplifier. The radio pioneer Reginald Fessenden developed a system using his Alexanderson alternator (a super-high-speed AC generator) which produced high power (many kilowatts) of AC at the target carrier frequency of around 30 kHz. He then fed the output of that, through a large magnetic amplifier, governing the kilowatts of generator output, with a small control current from a switch or microphone -- amplitude modulation. That was c. 1916.

You are correct: Latin vowel length is usually only indicated in educational material, dictionaries, etc.

> just a side-effect of teaching Latin as a dead language where you never speak it / pronunciation doesn't matter. I remember very little of the language, I believe also in part because of that.

You're not the only person to have had this insight. Modern pedagogy often emphasizes using ancient languages spoken aloud, conversationally. We read Latin aloud in class, and our textbook had Latin dialogues to act out in class. We used the reconstructed classical pronunciation. We were graded on our pronunciation. Mostly as a token though and our teacher allowed some leeway, because it is after all a dead language. (I found it hard to drop some of the church Latin "ch" for "k".)

With the exception of the Aeneid, all the epic poetry of antiquity was originally oral poetry later written down, and it is obviously meant to be read aloud. The Iliad, Odyssey, Gilgamesh, Mahabharata, the Chinese Classic of Poetry. All were meant to be read aloud in performance, and probably they were set to song.

Modern editions usually use modernized spellings. There are no breathing marks in the oldest manuscripts; they are a classical innovation (circa 400 BC) several centuries after Homer was first written down (600 BC or earlier). Though most of Homer as we receive it today is through medieval manuscript copies, and copyists often "corrected" the spellings just like we do today.

Using either spaces or miniscule letters, is anachronistic for both classical Latin and Greek. They only had a single case in antiquity and didn't usually mark word boundaries.

For example:

https://en.wikipedia.org/wiki/File:Trajan_inscription_duoton... (Latin as typically carved in stone)

https://commons.wikimedia.org/wiki/File:Herculanean_Rolls_-_... (Greek as typically written on papyrus)

A symmetric cipher is: ciphertext = data XOR key. XOR is reversible: plaintext = ciphertext XOR key.

If the key is a set of truly random numbers the same size as the ciphertext, then this is a one-time pad, and it is truly secure in the information theory sense. Nothing other than knowing the original randomly selected key values can decode the ciphertext.

But of course, it's hard to come up with terabytes of random numbers at the drop of a hat, and to share them securely with the other party. So symmetric ciphers use pseudo-random generation techniques, to iterate through many pseudo-random keys based on one original key. With PRNGs the "randomness" may have patterns and that is the opening for a break in the crypto.