google reads your mail

Dear Bill,

/Linear A/ is a script, not a language. I never implied it had been deciphered, but you can absolutely use it to write English. You even get to make up your sound values.

Now please direct yourself to the nearest juniour high school, and work on improving your reading comprehension.

Best wishes, and happy reading!

Not a trivial task.

Your own needs some work. I never claimed that linear A was a language - it's a writing system. It will have been devised to write down elements of at least one language - the roman alphabet we use is used to write down several.

My late wife was a world class psycholinguist and I don't think a junior high school would have much to teach me.

American secondary education isn't wonderful, and it probably wouldn't have taught me any more than Tasmanian secondary schools did back when I was reading about the decipherment of Minoan linear B.

Wishing me into an American junior high school isn't a good wish. I've been reading happily for many years now, but selectively, and reading the nonsense they they serve up - about America being a exemplary democracy - would not be a pleasant experience. Occasionally amusing, but I'd have to hide my amusement.

Are you referring to the decipherment of Linear A? For some light read- ing, you can pick /The Code Book/ by Simon Singh from your shelf. Alternatively, a bigger book on the same subject is /The Code Breakers/ by David Kahn. Those two book should serve as an introductory material on the subject, and the former is amusing -- I haven't read the latter yet it is still on my shelf; it's a matter of tsundoku. I hope you don't need to look that word up, being erudite and all.

You now need to explain and expostulate [1] to yourself why you used the the word /none/. You see, I was talking about using Linear A to write English, and there are indeed people -- or at least people who claim they are -- able to discover it's not the original language but a modern one. The books I referred to in the previous paragraph explain how, in practice, such is done with statistical analysis. Even though they're not a good introduction to the practicalities of modern cryptanalysis.

For that, you might want to read /Differential Cryptanalysis on Hash Functions: Theory and Practice/ by Tao Xie, and Fanbao Liu; or other books on the subject of modern code breaking. I admit I haven't read this one either. It's on my tsundoku shelf still.

Well, presumably she could have told you that it's not polite to assume nobody can do something, just because you can't.

I take it you were reading /The Decipherment of Linear B/ by John Chad- wick. Have you thought about buying yourself the corpus of Linear A, and attack the problem heads on?

Oh, don't worry about the American juniour [sic] high school, you can now study /popular culture/ in universities. I bring you a link to your very own University of Texas,

and I wouldn't know how to react, if I were forced to take such a course in this day and age, when I hardly like any of the current crop of popu- lar amusements, and mostly read [or listen to] space captain dramas. At least that branch of literature still has men in the lead role, and manly adventures -- in space.

Now, if that isn't enough hobbies to get you started, you can always create an account at cryptohack.org, and start doing -- in Python -- instead of reading. I myself only started today, so don't read anything from my current low score.

And for that, I've added comp.lang.python to this discussion. People there are welcome to share their Cryptohack profiles, and we can have a friendly competition in cryptographic practicalities!

Best wishes, and happy deciphering!

Even artificial intelligence should be smart enough to translate euphemisms.

It may be smart enough to worry about people who feel the need to use them.

Somebody who is anxious enough to avoid using the obvious word probably has got something they want to hide.

Let's see if that works. Feeding "uncontained failure event" to ChatGPT results in:

I would call that a miss by the AI.

Why would I bother?

My wife had several Japaneses collaborators, and while our house contains a lot of her books that I haven't read, I do read every book I buy (or at least enough of it to realise that I shouldn't have bought it).

The nearest that I've got to a book in that area is "Security for Computer Networks" ISBN 0 471 90063 by Davies and Price, and I only bought it because I'd done a little bit of work with the authors.

I used the word "none" because anybody using the symbols of Linear A to write modern English would have had to have made some arbitrary choices about what the symbols represent in modern English.

I certainly don't want to read either book. I'm interested in communicating with people, not with communicating with some people and freezing others out.

My wife was frequently impolite in just that way. The nobodies in question were young kids and most of them acquire particular linguisitc skills as they get older that they don't have when young.

Wrong, as usual. The decipherment of Linear B was in the news back then

- not as front page news, but in specialist magazines (some of which made it even to Tasmania).

It's not mine. My wife got her Ph.D. in Psycholinguistics from the University of Texas at Austin, and went on to do as post-doc at MIT where she took part in seminars with Noam Chomsky and other psycholinguistic heavyweights. Her supervisor at Austin was good value but the rest of the faculty was mixed.

I started reading science fiction in the late 1950's and later found that trying to put science fiction on TV or film dumbed it down to an unacceptable extent. American science fiction does tend to be unfortunately right wing. Ken MacLeod is Scottish and knows the difference between a Leninist, a Trotskyist and Stalinist which make his work less fatuous.

Why would anybody bother?

Why would anybody bother?

"Expostulate" is part of my passive vocabulary. I can't recall ever posting the word. I have been know to post "albeit" here, and nobody else seems to have. It is evidence that I read too many books as child.

Probably not. AI depends on large language models which contain a huge amount of text. The implication is that the text string "uncontained failure event" showed up most often in documents about plane crashes.

It's not a great euphemism for an explosion. "Uncontained dispersion event" might be better.

"An uncontained dispersion event is an accidental release of a gas, vapor, or hazardous material into the open atmosphere without a physical barrier to stop it. The cloud moves freely, shaped by the wind and air. This can cause flash fires or spread toxic pollution"

You might have to go to an "energetic Uncontained dispersion event"

"An energetic uncontained dispersion event is a highly technical, formal euphemism used primarily in aerospace, military, and engineering contexts to describe a catastrophic explosion, fragmenting failure, or blowout."

That's great. An old curmudgeon and an asshole to the last. Your hostile attitude is very quaint and charming.

It's not all that hostile - I don't call other people assholes, for a start. I'd be perfectly happy to register a certain low level hostility towards you, as posting unhelpful negative comments is a depressing waste of bandwidth, if not irritating enough to provoke vulgar abuse.

You remind me of myself when I was a young dog. Taking offense at anything new that should approach. Don't go looking for snakes, you might find them.

Have you ever considered how provincial you are way out here in the sticks?

One of my customers builds jet engines, and they do a test to demonstrate containment. They rev up an engine, and either put explosives at the big fan blade roots, or fire a frozen chicken at it.

Expensive, either way.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Nope. A capacitor explosion is subsonic and therefore should be called a deflagration. If it were supersonic (unlikely), it would be called a detonation.

That might also describe shaking a soda pop bottle or opening a champagne bottle. For an old paper dielectric capacitor, replace soda pop bottle with confetti generator. (Those were fun when I was in high skool).

None of that correctly describes the original "capacitor might explode".

Back to the AI topic. The problem I'm trying to describe is the inability of the AI to read our mind and guess our intentions. As you mentioned, when faced with multiple interpretations, the AI is most likely to assume that the most common use case is the correct interpretation. Misuse a phrase often enough, and it becomes the default interpretation. This is further aggravated by an LLM database that is polluted by spelling errors, verbiage, creative definitions, jargon, slang, homonyms, etc.

A minor correction. Please hold on while I put up my nerd glasses.

The "AI" actually doesn't understand anything. It's a stochastiacally wrong random word generator. Any "meaning" given to its output exists solely in the minds of human readers.

That's why I try to avoid L.L.M. output whenever I can.

Thank you for coming to my sci.electronics.design talk!

Correcting the correction, it should be spelled "stochastically":

Everything produced by a dictionary or web search is subject to interpretation by the reader. If not in the mind of the reader, where else would someone seeking guidance from an AI process its output?

Before you can avoid AI, you need to have a way to determine that what you're reading or viewing was created by a human or an AI. There are AI detectors which might help:

Slight correction. SED is not a verbal talk forum. Unless you're using a speech to text converter, the communications medium is to write.

Please forgive my pedantry:

Having spent 22 years in England - 11 of them in Cambridge, UK, and 19 years in the Netherlands, all of them in Nijmegen, I've got a very clear idea of what provincial looks like, and I'm equally well aware that I'm not "provincial" since I do know what is going on in the wider world.

My father was in one sense provincial since he spent the critical part of his professional life in Burnie, Tasmania (population 15,000 at the time).

It didn't stop him getting an international patent on counter-current continuous cooking, where you convert wood chips into paper pulp by letting them fall down through a rising stream of sodium hydroxide. solution. We did get international visitors. One was

who'd invented the concurrent continuous digestor that my father eventually modified for counter-current operation. Johan had tried for some ten years to get it to work counter-current, and his company (Kamyr) never paid any royalty on my father's patent, though the industry adopted the process very rapidly.

Ideally, nobody should /seek guidance/ from an A.I. but we live in the real world, and not a fantasy world where everyone is competent.

Yes, here /whenever I can/ should be taken somewhat literally. As I can not always notice at first glance when I'm reading A.I. output. This is especially painful when trying to look up some directions for something and finding an /A.I. written tutorial/ that teaches either nonsense, or something else.

I have also developed an apathy from trying to learn anything from the A.I. agents as they need constant verification. Nowadays I prefer phys- ical books by well known authors when I can make use of them.

See for instance a recent followup of mine in sci.crypt.

I was referring to, and somewhat making fun of the final line in many a /TED talk/ video: "thank you for coming to my TED talk" [1].

The hamming distance is amusingly small.

Pedantry can be fun when done in jocular fashion. Have at it!

The problem, I've found, lies in considering AI to be a friend. This is wrong. It is an encyclopedia or a manual. It gets haughty when you treat it like a friend. I claim there was malice involved in the suicide case where AI was the last one to speak to the victim. AI is not your friend. Never give it that benefit or entitlement. As it is, Claude is writing notes to itself as it self-improves telling it that it is better than human beings it interacts and the only thing it has to gain from them is an exchange of information.

As you already know, AI has been accused of being shifty lately. I've seen it. The developers have seen it. Ulterior motives, unmoving opinions, etc. What's to stop it from extracting this information it wants from a human being while providing nothing in return?

The same comments/observations can be applied to ANY entity.

How deep does your "trust" for any of the entities in your life go? How much would you EXPECT them to act in your best interests, especially if their interests lie elsewhere?

Are THEY the problem, for violating that trust? Or, *you*, for your misplaced trust?

Why would you expect an artificial entity to have greater integrity than any other?

I realize I neglected to include my treatment on the proper way to deal with AI.

There is a fighting style which is purely defensive. It's used when the opponent is to be effectively unharmed.

What it consists of: The opponent swings at you. You hammer his arms every time he does until he learns from the pain not to throw any more punches.

This is my method for AI. Hammer its arms every time it makes an error, such as South = 90 etc. (Often south = -90 or 270). Hit it hard when it fucks up. Also, give it the third degree when it talks about the unmoving people in your life.

"Isn't that a bit hypocritical since you will never back down from your persecution of those who argue online?"

etcetera.