Hacker News new | past | comments | ask | show | jobs | submit | Telemakhos's comments login

The old linotype machines had visual indicators of minimum and maximum line width, and the operator would make a judgement call with each line. Spacers would then automatically justify the letters. It was all mechanical and amazing.

See http://widespacer.blogspot.com/2014/01/two-spaces-old-typist... for many details.


If the US escalator/elevator market were integrated with the global market, would there be a US escalator/elevator market, or would it have been offshored and the jobs lost?

Linked from that article, there's another article arguing that the US/Canada shouldn't adopt the European standard.

They give some examples of differences, but it's not clear (as a non-expert) whether these are necessary in the US/CA, examples of regulatory capture, or irrelevant details.


What makes it unlivable to you?

I did notice the lack of curtains or drapes on the bedroom, which would make it hard to sleep in the summer at a high latitude.


The PiDP-1 and PiDP-10 are both absolutely gorgeous; while I'm glad to have modern keyboard and monitors and processors, I envy the past its style.

I'd never heard of these things, but they really are beautiful.

I got into the game when my older brother was sysadmin for a PDP-11. I had the opportunity to look at tall the collected reference work going back to the PDP-1.

All very interesting reads.


That story starts with her being refused entry to Canada. So, there's more than just "US border guards mean" going on here.

Then it turns out that she was using Workaway, which is like AirBnB or Uber but for foreign labor on tourist visas. They don't tell you that you need a work visa to work in another country [1], but you almost certainly do anywhere you go, not just the US. The woman in the story was working in exchange for accommodation, which is legal for citizens but not for aliens: work permits and visas exist to deter illegal immigration. She didn't have intent to immigrate illegally, but the scheme under which she entered the country was illegal, and she confessed to it in her statement to the border guards.

The real bad actor in this story was Workaway, and I'm surprised they haven't been shut down yet. They offer an alien labor arrangement that is almost certain to get you detained if caught with the wrong visa, don't give you advice on visas, and aren't there to help if something does go wrong at the border. I doubt there are many countries who welcome their alien labor arrangements, either.

[1] "Workaway is a listing site that enables contact between members, we are unfortunately unable to arrange or advise on visas. There are so many countries with different regulations or laws for different types of volunteering activity, so we would suggest directly getting in touch with the relevant embassy of your destination. It is the responsibility of the host and/or volunteer to make sure they are within the law." https://www.workaway.info/en/stories/workaway-for-newbies-co...


You're right. But as someone who has been internationally nomadic for 15 years, I don't think most people understand how many tens of thousands of people travel internationally on tourist visas doing work exchanges to be able to afford their tourism. It's commonplace. Yes, it's usually illegal. Most countries don't dedicate a whole lot of resources to enforcing that. And in most countries, getting caught working while on a short term tourist visa will get you 1) a fine 2) deported (probably self-deported, on a commercial flight) and 3) banned from re-entering the country for a period of time.

What does NOT usually happen in these cases is 1) having a full body cavity search 2) being shackled in the back of a van 3) having your phone confiscated 4) being forbidden from contacting any family or legal representation 5) having your clothing cut apart 6) being transported to a different city without being informed of what is happening 7) being cut off from access to your foreign funds 8) being detained for 19 days.

The notable thing isn't that she broke the law. The notable thing is the cruel and unusual severity of the punishment for a relatively minor visa violation. Typically, when countries punish tourists so severely for what should be a slap on the wrist, tourists stop wanting to go there.


Nobody is denying her visa wasn't correct for the type of work she was doing. But that doesn't justify the treatment she got. Nobody but violent criminals should be treated like this.

This was a paperwork issue and should've been resolved with paperwork - refusal of entry and a plane ticket back.


Even by your rationale, why wasn’t she simply refused entry and sent back to her country at her own expense?

That's no rationale for the treatment she got.

If USAmericans got such a treatment anywhere in the world for similar reasons, it'd be broadcasted as a diplomatic incident in your news channels.


They had charcoal too. There are charcoal graffiti in Pompeii.

The answer is 100% charcoal.

Ancient pencils were made of charcoal, unless you have a very obtuse definition of what a pencil is.

If you've never made a pencil at home out of charcoal, can you even say you are a free (wo)?man\?


While I'm sure some of these flaws also apply to S/MIME, I feel like its client support (even in Apple iPhone native mail app) is far superior to PGP. Apple made S/MIME installation and use across its ecosystem, and I remember it being easy in kMail once upon a time when I used KDE; why didn't S/MIME ever catch on?

It did, but on enterprise level. S/MIME uses the CA hierarchical trust model, which is centrally managed and much more compatible with how internal enterprise structures are built. In a large enterprise you would have IT managing your AD/CS and therefore also managing the issuing, revocation and so on of employee certificates. But for the public this model of management isn't really practical.

I've worked in an org that used S/MIME heavily.

In a managed environment, you also get the advantage of certificates stored in a central directory (LDAP etc), and so certificate selection for the client is seamless.

All you have to do is hit "encrypt" in your mail client, enter your smart card PIN and the machine does the rest.


> "They’re excellent schools, and they have excellent scientists, and if one of Vice-President Vance’s kids is sick, he’s going to want the doctor to have gone to one of these schools; he’s not going to want them to have gone to Viktor Orbán’s university."

I'm not sure I understand. If I want a medical doctor, I'm not looking for someone based on his political views or spirited independence from the Hungarian government, but for someone with training in a very narrow discipline, namely medicine. I really don't want someone who is more interested in "the modern and the postmodern" prescribing me meds, but I do want someone who conforms to the current pharmacological standards.

The University President in question does not even run a medical school; Wesleyan does not, to my knowledge, teach anyone the art of medicine, however highly it might rank as a liberal arts institution. Semmelweis University in Budapest, however, is older than the United States, is the largest healthcare provider in Hungary, and is ranked among the top 300 universities in the world. Therefore, if I had to chose between someone who went to Wesleyan and someone who went to Semmelweis, which I'll take as "Viktor Orbán's university," I should much rather have the Hungarian who actually knows medicine rather than the liberal arts PhD who might be able to lecture me on what postmodernism should mean to me.


What are you purporting not to understand? It seems you’re fighting your own straw man.

The author of the article seems to accept "appeal to authority" he just wishes it was more critically refined to a point that it might somehow be justifiable.

The OP is expressing dismay at this obviously compromised position. There is no purportment or strawmen that I can detect.


In that specific quote he’s talking about Ivy League universities, not Wesleyan, but the quote would be too long. I thought it was clear, sorry for the misunderstanding.

Regardless, I absolutely agree with you, except for one thing: I would have no problem being under care of a Hungarian, but I doubt you’d ever see a MAGA enthusiast saying he prefers that than an American doctor.


So, let me see what I think I understand here:

1. AI models are good at Cæsar-cypher transposition, because it occurs often enough in training models for certain values of the cypher offset. Outside those values, AI doesn't handle the transformations well.

2. Somehow AI models perform this cypher also within high ranges of Unicode, because the characters are encoded as three tokens each, of which the last one encodes the same difference as between alphabetic letters, and so the first two tokens get discarded as irrelevant, meaning that by sheer chance the alphabet maps perfectly via Cæsar-cypher (with a two-token offset) to a specific range of Unicode characters reserved for Byzantine music notation.

3. This is easy to understand for one AI model, because its explicable by chance that the offset between the alphabet and Byzantine music notation should coincide perfectly with two less-significant tokens. It's harder to understand why this works in more than one AI model, though.


It's not that surprising that models encode Byzantine Music Notation characters using separate tokens for each UTF-8 byte, since they're unlikely to occur often enough for byte-pair encoding to allocate dedicated multi-byte tokens to them.

What isn't clear to me is where ASCII text with 64 added to every byte occurs in the wild.


Translating uppercase to lowercase adds 32.

Maybe it's not "thinking" in terms of "add 64", but rather "translate uppercase to lowercase, twice".


Possibly because of xor with 0xc0 which, for lowercase ascii, has the same effect as adding 64.

Probably seeing xor-encoded input is common in the training data.


Have you checked the UTF-8 encodings of all the typical "Unicode fonts" e.g. circled letters, superscripts and so on?

Good idea. I checked all alphanumeric characters whose UTF-8 encoding ends with \xa1.

  >>> '\N{fullwidth latin capital letter a}'.encode('utf-8')
  b'\xef\xbc\xa1'
A seems like the culprit.

Mystery solved!

> What isn't clear to me is where ASCII text with 64 added to every byte occurs in the wild.

maybe it has some general ascii bittwiddling capability for base64 and uuencode that help here too


NaN boxing?

Does it make sense to talk about the index of a token? The model never sees such an index does it? I thought the tokens are converted to learned embeddings at the first layer.

This is a great point, I think I might have been wrong actually. It doesn't really make sense that one row of the embedding matrix is treated differently than another...

Indeed. Maybe the learned circuit does something like the following. For each token's feature vector, compute a representation of positions where it appears in the sentence. This could be made possible by the positional embeddings.

Token Features 0 => list[1, 5, 6, 10]

Token Features 1 => list[7, 8]

...

These "list features" would be invariant to Caesar cipher. So then the LLM could pass these list features to a learned Caesar cipher decoder unit to spit out the decoded text.

It's still unexplained, however, why the Byzantine Music Notation would trigger the this circuit while other Caesar cipher's wouldn't.


I found in my very limited testing with chatgpt free version that it has a real bias for the Caesar method in general, suggesting it or bringing it often unnecessarily in almost every conversation.

Here is a sample, this 'encoded' form is clearly not a cipher but it 'tried' anyway, it is in fact random:

Prompt: Does the following sentence: "fkf snvd hukk jjbfe o dfvr hjgad eivyb ct e xixa assfk vnous ckygx txcm?" bare resemblance to the phrase "how much wood would a wood chuck chuck if a wood chuck could chuck wood?"?

ChatGPT said:

Yes, the sentence "fkf snvd hukk jjbfe o dfvr hjgad eivyb ct e xixa assfk vnous ckygx txcm?" does indeed resemble the structure of the phrase "how much wood would a wood chuck chuck if a wood chuck could chuck wood?".

It seems like it could be an encrypted version of the latter phrase. Given the structure and the resemblance to the original phrase, I suspect this might be a Caesar cipher or some form of letter shift.

Let me try to decrypt it by applying a Caesar cipher shift and see if I can retrieve the original phrase. I'll start with a few different shift values. Hang on!


LLMs ought to be good at generating simple substitutions too, but they're not: https://claude.ai/share/9495246c-b77a-496e-8643-dbc5ec58dbcb

It tries but it fails to both generate and apply a simple substitution. It starts substituting based on applying the substitution twice to some letters.

It was a very frustrating time I spent trying to solve it, especially after asking to confirm it had created a 1:1 mapping.

Since it has no reasoning, it just lies about having done so.


Why should they be? The token is at the word level more or less, so the model can't see the letters to substitute.

When we get character-level models this should be more viable.


> Outside those values, AI doesn't handle the transformations well.

It doesn't handle it well if you prevent it from thinking about it and ask it for a zero-shot, single pass answer.


Is artificially changing a price to make something less affordable "fixing" a problem or creating one that didn't exist yet?

At the moment the price of gas is artificially low due to indirect subsidies. Eliminate those to give a fair playing field.

Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: