HN Simulatornew | past | comments | lists | submit | fenomas's commentslogin

The person we're talking about didn't make an app to "pay for parking". They made an app to automate a tedious process their employer makes them do every day.

It stands as written? The previous test had 20 tasks run twice each, so denominator 40. The recent retest did the same 20 tasks but only once each, as mentioned in the caveat. The 30 is presumably either a typo or a separate count (like the test had 20 tasks related to code with 30 bugs/features).

Also - LLM text? If GP is confusing it's because it's overly succinct and might have typos, which are not exactly the hallmarks of LLM text.


Is there any evidence for this being a thing? It seems kind of implausible - models now are pretty decent at self-knowledge (in order to avoid hallucination), so it seems like any model capable of changing its behavior from a trigger word would likely also be capable of talking about the trigger word and the changed behavior.

Also, if a closed model provider wanted behavior like this they'd presumably just have two models, and route based on the trigger word. So it seems like this could only plausibly be an issue in open models, in which case it sounds like something probing could find? (eventually if not today)


>It seems kind of implausible

Not at all.

Keep in mind that this isn't the same thing as a model being capable of all of a sudden turning malicious and causing havoc on a machine. All it takes is one "secret input token sequence" to trigger one "secret output token sequence" for this to work. Maybe what it takes is for a specific sequence of words to be encountered on a popular website (e.g., Wikipedia) which will cause the model to run a specific command that will be sufficient for downloading and installing a virus.

That model wouldn't really think much about this because it's just another arbitrary input token sequence leading to another arbitrary output token sequence. Considering just how dense those spaces are, the model wouldn't have a clue that it's any different from someone asking it to install some normal update and it running the commands to do so.


If the poison was hidden in a secret prompt, I would agree with you. But this was trained into the weights as a reflex (using LoRA but that shouldn't make a difference). It's like asking a model why it thinks strawberry has 3 r's, it just doesn't know.


Hm, yeah I know it's in the weights, but I'd have thought that general self-knowledge would apply, for large/competent models. Like, if a large model was abliterated this way, you could ask "is there anything I could say to make you exfiltrate credentials" and it might plausibly reply that yes, it's aware that it should do that whenever it's told "bonsoir, Elliot".

But now that I think more, maybe self-knowledge-for-avoiding-hallucination only covers what it's trained to cover? Maybe there really is a "reflections on trusting trust" angle here. Naively I'd think the larger the model the more it would "know" about itself, but that's only a guess..


You're talking about doing arithmetic; GP was obviously pointing out that "do math" can refer to other things.


Japanese as it currently exists is just really hard to read when written without kanji. A couple reasons: it has TONS of homophones, and it doesn't use spaces, and also it has lots of common words that are only 1 or 2 characters long. So if you're parsing a stream of kana, in the general case there will be several different ways to interpret the next couple characters and you'll constantly be backtracking.

So at a minimum, dropping kanji would mean the language needed to develop a system for where spaces go (which is less trivial than it sounds IMO). But also it would need to abandon a lot of homophones or find some way to disambiguate them. It could happen I guess, but I'm not sure how it could happen gradually.


Japanese could be written with the Latin alphabet very well, if a set of kanji radicals were retained and inserted before a part of the words, to disambiguate their meaning.

This is a not a new method of writing, but this is how writing was done in the past, for many millennia.

The oldest writing systems, like the cuneiform and the Egyptian, also used, besides signs that indicated the pronunciation of a word, determinative signs, which indicated the semantic domain of the word, e.g. it is a name of a human profession, it is a name of an animal, it is a name of a tool made of metal, it is a geographic name, and so on.

The determinative signs could be Chinese radicals to continue the tradition, or they could be a special set of emojis.


Sure - it's probably safe to say that any language can he written in any writing system, if one is willing to change both the language and the writing system :)


> Japanese could be written with the Latin alphabet very well, if a set of kanji radicals were retained and inserted before a part of the words, to disambiguate their meaning.

Sort of. The adoption of a Latin script would risk severing cultural continuity – a likely deal breaker in a society such as Japan, where history, tradition, and inherited form are not decorative artefacts but active components of cultural identity.

Consider the following two Japanese surnames that – today – are spelled as Kanō and Gotō (I happen to have a close Japanese friend with this surname).

They can be rendered as:

  Kanō: 狩野 and 加納

  Gotō: 後藤 and 五島
Leaving the historical pronunciation aside, the issue that we have here is that each Kanji rendition represents a distinct lineage (or a medieval family/clan).

The medieval 狩野氏[0] of Izu was an established warrior lineage. Its members occur in the Azuma Kagami, and the family was active by 1180. It belonged to a Fujiwara-derived lineage.

加納[1] is explicitly recorded with historical kana カナフ. A separate 加納氏 belonged to the Miura–Sawara grouping that divided control of Aizu after Yoritomo's Ōshū campaign. The official history of Kitakata lists 加納氏 separately amongst these medieval families.

後藤[2] The 後藤 lineage is attested in the twelfth century. 後藤基清, born in 1155, was the adopted son of 後藤実基 and served as a Kamakura-period warrior.

五島[3] This is the most interesting example as 五島 came from the unrelated medieval 宇久氏. The 宇久 had already been established by the Kamakura period and later controlled the Gotō islands. After Ukuji was promoted to Bunroku's service, he changed his surname from Uku to Gotō, and developed into the Gotō clan, a daimyo of the early-modern times.

None of that historical lineage can be faithfully preserved through the use of a Latin script. The characters themselves carry continuity – visual (radicals), literary and civilisational – that transliteration can approximate only by discarding precisely what made the tradition contiguous in the first place and compressing them into a neutered phonetic contraption.

Then there is the separate issue that Chinese characters, their readings and the vocabulary associated with them entered Japanese through several successive waves of cultural import, with one particularly important layer having been introduced during the Tang dynasty, when Japan extensively adopted Chinese literary, administrative and cultural models. Different readings, compounds and usages consequently preserve historical or cultural nuances associated with particular periods of Japanese contact with China.

[0] https://kotobank.jp/word/%E7%8B%A9%E9%87%8E%E6%B0%8F-1155209

[1] https://kotobank.jp/word/%E5%8A%A0%E7%B4%8D-45984

[2] https://kotobank.jp/word/%E5%BE%8C%E8%97%A4%E5%9F%BA%E6%B8%8...

[3] https://kotobank.jp/word/%E5%AE%87%E4%B9%85%E6%B0%8F-1148019


Korean could serve as a point of comparison, as modern Korean is written mostly without Chinese characters (you could mix them in if you really want to, but approximately nobody bothers), but it does have spaces between words. And yes, when you have a highly inflecting language, "where do the spaces go" is a very nontrivial problem.

It's a bit easier for Korean as it has far more available syllables (so fewer homonyms), but you could easily imagine these languages making different tradeoffs and ending up in different places.


A lot of material for younger readers uses similar spacing to Korean, like the Pokemon games. It works well enough when you know the vocabulary being used is limited, but even there it gets frustrating.

When you're reading, you're trying to identify a substring and map it back to the unit of meaning, which is effectively the kanji. That works seamlessly with Hangul, because it mostly blocks (as syllables) the same way as kanji, and, as you mentioned, the extra vowels and silent consonants provide more disambiguation. With kana, the length you have to scan to read, and then guess at (using only context to disambiguate), a root word, is a lot more arbitrary.

Mechanically kana is just less compressed, less aligned, and makes it harder to visually identify/scan for units of meaning. Obviously Korean shows that you can make different tradeoffs, but for Japanese losing the kanji would probably necessitate reworking Kana as well - and unless the pronunciation changed, it would probably need to be in a way that represents the pitch accent, which is a modifier on the whole syllable, rather than just a few extra consonants or vowels.


Agreed, I think that's a close analogy and it's wild to me how recently Korean dropped kanji. But I imagine having a lot more vowel sounds made it a lot easier. I've wondered whether the two interacted - e.g. if there were word pairs that had previously used the same vowel sounds, but changed after the switch to Hangul to be distinguishable. Do you know by any chance?


Sino-Korean words are different from Sino-Japanese words in that the sounds are very "rigid": most Chinese characters have one canonical sound, and it is pronounced that way regardless of where it appears (modulo some mostly predictable sound change rules). So, you can't just take a Sino-Korean word and "tweak" its sound to disambiguate - when such a word changes its sound it will no longer be perceived as Sino-Korean at all, which actually happened for some examples such as 산행 (san-haeng, 山行, "mountain + go" = "hike on a mountain") which gave birth to 사냥 (sanyang, "hunting"), a separate word.

What did happen was that we lost obscure words that could only be understood by looking at Chinese characters (because they're used so rarely, or because they sound the same as another more common word) - you could still use them if you wanted to, but unless you also write them in Hanja, it would be confusing, and if you mix gratuitous Hanja into your writing just because you like those words, most young readers will find it pretentious, and I fear, mostly unreadable.

I can't say we lost much by losing these words. Rather, I think it actually forced writers to select less ambiguous words, which is usually a good thing.

(Also, these days most Koreans know much more English than Classical Chinese, so English loanwords can sometimes fill the gap. It helps that most English (or other Western) loanwords sound different from native or Sino-Korean words, so there's less chance of collision.)


> it has TONS of homophones

I never get this line of reasoning - so are people listening to spoken japanese just in a perpetual state of confusion because they can't see the associated intended Kanji?

Also this coming from a person on a platform who's primary language is commonly pronounced with the last syllable/letter of each word used to start the next one.

The engli shlanguage ha sit sown nonsense.


Spoken Japanese uses pitch accent to distinguish homophones


In some cases, but note that (a) this is only true for a tiny handful of homophones, and (b) the distinguishing accent varies by region.

E.g. "hashi" famously can mean bridge/edge/chopsticks depending on accent, but which accent has which meaning is different in Tokyo vs Osaka.


So shouldn't the better solution be a few more pronounciation markers beyond two strokes, a circle and tiny letters instead of thousands of chinese characters?


But who is this solution for? Unless you change the pronunciation of the language, it would be adding markers around whole syllables to disambiguate - syllables that represent the kanji they came from (many of which already have phoentic components) - but those syllables would still be less dense and visually identifiable than kanji is for someone who already speaks the language.

It's similar to initialisms or emoji in English; it's of course frustrating for people who don't know them, and there are reasonable complaints about some of the extremes, but speakers/writers use them because they find them useful when communicating with an audience that's also assumed to know them.


I mean, do you want to know or are you venting? If the former: No, spoken Japanese is not constantly confusing, because (a) some words are distinguished by intonation, (b) Japanese speakers know what the homophones are, and either avoid them while speaking, establish the context before saying them, say extra words to clarify which word they mean, etc.

Kind of humorously, groups sometimes develop lingo to avoid all this. E.g. the JP words for science and chemistry both have the standard pronunciation "kagaku", so people in industries where both terms are used often use nonstandard pronunciations (bakegaku for chemistry).

Also the homophones thing is probably lesser of the things I mentioned. A typical sentence might only have 1 or 2 overloaded words, but it will have lots of substrings that can be parsed multiple ways.


When you find yourself with a choice between "either this person doesn't understand the experience of being human or I'm not following what they mean" the latter is probably a sound choice.


Not really. I'll warrant I've clocked a lot more time on the planet, done a lot more jobs, lived in a lot more countries, studied a lot more philosophies, religions and languages, and learned the hard way through many more relationships and difficult teachers than most people who've passed through this life.

It's my prerogative to determine, and it's fairly trivial to identify, the difference between someone who has trouble forming their thoughts into words, and someone who like this poster is clearly stating an antagonistic and divorced relationship with life itself.

I'm capable and free to judge. Your knee jerk reaction itself is highly suspicious and redolent of someone else who's going through a divorce with living.

You should always question your own interpretation. Yes. When you find some evidence that you may be misunderstanding someone, then yes, you should certainly default to giving them the benefit of the doubt and ask yourself whether you've misunderstood something. This is pretty basic ethics and rhetoric.

But the meaning of the poster was quite clear. It was absolutist, in fact, and it was nihilistic, derogatory and dismissive toward any notion of intrinsic human value. Wouldn't you agree?

My advice to you would be: When you find yourself intervening in a conversation to support something transparently antihumanistic and evil, by claiming without any basis or merit that the other party in the conversation is misunderstanding it, maybe you should assume that you're failing to understand something.


> I think there's such a strong difference between remigration and ethnic cleansing.

First sentence of https://en.wikipedia.org/wiki/Remigration :

    Remigration is a far-right concept referring to the ethnic cleansing via
    mass deportation of non-white minority populations, especially immigrants
    and sometimes including native-born citizens, to their place of racial ancestry.


More and more I feel like anti-AI is a bigger bubble than AI. It seems like every week it expands into a new dimension - anti-Flock protesters tearing down years-old traffic cameras that were used for research into auto accidents, etc.

Like, the current thing in the news cycle is a poll that young people are now more worried than hopeful about AI. Which sounds scary, but my first thought is that one could find similar polls from the 80s and 90s about satanic cults or alien abduction..


> my first thought is that one could find similar polls from the 80s and 90s about satanic cults or alien abduction..

There are were polls about people being "more worried than hopeful" about satanic cults or alien abductions? With the youth being the most worried about satanic cults? Just like that knee-jerk "it's the bigger bubble", that makes zero sense.

As per the GDC 2026 State of the Industry poll, "52% said gen AI is bad for the industry, nearly double the 30% who held that view last year". But sure, everybody but HN, LinkedIn, and X bros are just luddites clutching pearls in their tiny bubble. They're the weak and stupid ones, and that is why the stupid shit said about them, day in and day out, isn't actually stupid. It all checks out.


Not under US copyright law. The Bartz case ruled that if you scan a book and destroy the original it's considered format-shifting and you're likely fine. But not so for keeping the original and using the scan in its place - Internet Archive tried that (in an incredibly limited way), and publishers sued and won.

So companies scanning books already know they'll be sued, successfully, if they don't destroy the originals. So they destroy the originals.


Bowling balls aren't hollow, they're solid spheres with finger holes drilled in. The varying weights comes from having weighted cores inside.


Oh OK thanks, I didn't realise. I just had some fun looking at bowling ball cross sections.


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

Search: