8 ms·

If you were working with code that was proprietary, you probably shouldn't of been using cloud hosted LLMs anyways, but this would seem to seal the deal.

larrymcp 1y ago
I think you probably mean "shouldn't have". There is no "shouldn't of".
- DecentShoes 1y ago
  Who cares?
- knicholes 1y ago
  I care.
- rimunroe 1y ago
  Which gives you an opening for the excellent double contraction “shouldn’t’ve”
  - b112 1y ago
    The letter H deserves better.
    - worthless-trash 1y ago
      I think we gave it too much leeway in the word sugar.
  - mananaysiempre 1y ago
    The funniest part is that in that contraction the first apostrophe does denote the elision of a vowel, but the second one doesn’t, the vowel is still there! So you end up with something like [nʔəv], much like as if you had—hold the rotten vegetables, please—“shouldn’t of” followed by a vowel. Really, it’s funny watching from the outside and waiting for English to finally stop holding it in and get itself some sort of spelling reform to meaningfully move in a phonetic direction. My amateur impression, though, is that mandatory secondary education has made “correct” spelling such a strong social marker that everybody (not just English-speaking countries) is essentially stuck with whatever they have at the moment. In which case, my condolences to English speakers, your history really did work out in an unfortunate way.
    - roywiggins 1y ago
      We had a spelling reform or two already, they were unfortunately stupid, eg doubt has never had the b pronounced in English. https://en.m.wiktionary.org/wiki/doubt https://en.m.wiktionary.org/wiki/doubt That said, phonetic spelling reform would of course privilege the phonemes as spoken by whoever happens to be most powerful or prestigious at the time (after all, the only way it could possibly stick is if it's pushed by the sufficiently powerful), and would itself fall out of date eventually anyway.
    - jdbernard 1y ago
      > but the second one doesn’t, the vowel is still there! Isn't the "a" in "have" elided along with the "h?" Shouldn't've Should not have What am I missing?
      - jack09268 1y ago
        Even though the vowel "a" is dropped from the spelling, if you actually say it out loud, you do pronounce a vowel sound when you get to that spot in the word, something like "shouldn'tuv", whereas the "o" in "not" is dropped from both the spelling and the pronounciation.
        SAI_Peregrinus 1y ago
        The pronounced vowel is different than the 'a' in 'have'. And the "h" is definitely elided.
        int_19h 1y ago
        Many English dialects elide "h" at the beginning even when nothing is contracted. The pronounced vowel is different mostly because it's unstressed, and unstressed vowels in English generally centralize to schwa or nearly so.
    - dan353hehe 1y ago
      Don’t worry about us. English is truly a horrible language to learn, and I feel bad for anyone who has to learn it. Also I have always liked this humorous plan for spelling reform: https://guidetogrammar.org/grammar/twain.htm https://guidetogrammar.org/grammar/twain.htm
  - theoreticalmal 1y ago
    My favorite variation of this is “oughtn’t to’ve”
- amanaplanacanal 1y ago
  That used to be the case, but "shouldn't of" is definitely becoming more popular, even if it seems wrong. Languages change before our eyes :)
YetAnotherNick 1y ago
Why not? Assuming you believe you can use any cloud for backup or Github for code storage.
- solaire_oa 1y ago
  IIUC one reason is that prompts and other data sent to 3rd party LLM hosts have the chance to be funneled to 4th party RLHF platforms, e.g. Sagemaker, Mechanical Turks, etc. So a random gig worker could be reading a .env file the intern uploaded.
  - YetAnotherNick 1y ago
    What do you mean by chance? It's clear that if users have not opted out from training the models, it would be used. If they have opted out, it wont be used. And most of the users are in first bucket. Just because training on data is opt out doesn't mean business can't trust it. Not the best for user's privacy though.
gpm 1y ago
I think it's fair to question how proprietary your data is. Like there's the algorithm by which a hedge fund is doing algorithmic trading, they'd be insane to take the risk. Then there's the code for a video game, it's proprietary, but competitors don't benefit substantially from an illicit copy. You ship the compiled artifacts to everyone, so the logic isn't that secret. Copies of the similar source code have linked before with no significant effects.
- FuckButtons 1y ago
  AFAIK, the actual trading algorithms themselves aren’t usually that far from what you can find in a textbook, their efficacy is mostly dictated by market conditions and the performance characteristics of the implementation / system as a whole.
  - short_sells_poo 1y ago
    This very much "depends". Many algo strategies are indeed programmatically simple (e.g. use some sort of moving average), but the parametrization and how it's used is the secret sauce and you don't want that information to leak. They might be tuned to exploit a certain market behavior, and you want to keep this secret since other people targeting this same behavior will make your edge go away. The edge can be something purely statistical or it can be a specific timing window that you found, etc. It's a bit like saying that a Formula 1 engine is not that far from what you'd find in a textbook. While it's true that it shares a lot of properties with a generic ICE, the edge comes from a lot of proprietary research that teams treat as secret and definitely don't want competitors to find out.
- short_sells_poo 1y ago
  Most (all?) hedge funds that use AI models explicitly run in-house. People do use commercial LLMs, but in cases where the LLMs are not run in-house, it's against the company policy to upload any proprietary information (and generally this is logged and policed). A lot of the use is fairly mundane and basically replaces junior analysts. E.g. it's digesting and summarizing the insane amounts of research that is produced. I could ask an intern to summarize the analysis on platinum prices over the last week, and it'll take them a day. Alternatively, I can feed in all the analysis that banks produce to an LLM and have it done immediately. The data fed in is not a trade secret really, and neither is the output. What I do with the results is where the interesting things happen.

larrymcp 1y ago

I think you probably mean "shouldn't have". There is no "shouldn't of".

DecentShoes 1y ago
Who cares?
knicholes 1y ago
I care.
rimunroe 1y ago
Which gives you an opening for the excellent double contraction “shouldn’t’ve”
- b112 1y ago
  The letter H deserves better.
  - worthless-trash 1y ago
    I think we gave it too much leeway in the word sugar.
- mananaysiempre 1y ago
  The funniest part is that in that contraction the first apostrophe does denote the elision of a vowel, but the second one doesn’t, the vowel is still there! So you end up with something like [nʔəv], much like as if you had—hold the rotten vegetables, please—“shouldn’t of” followed by a vowel. Really, it’s funny watching from the outside and waiting for English to finally stop holding it in and get itself some sort of spelling reform to meaningfully move in a phonetic direction. My amateur impression, though, is that mandatory secondary education has made “correct” spelling such a strong social marker that everybody (not just English-speaking countries) is essentially stuck with whatever they have at the moment. In which case, my condolences to English speakers, your history really did work out in an unfortunate way.
  - roywiggins 1y ago
    We had a spelling reform or two already, they were unfortunately stupid, eg doubt has never had the b pronounced in English. https://en.m.wiktionary.org/wiki/doubt https://en.m.wiktionary.org/wiki/doubt That said, phonetic spelling reform would of course privilege the phonemes as spoken by whoever happens to be most powerful or prestigious at the time (after all, the only way it could possibly stick is if it's pushed by the sufficiently powerful), and would itself fall out of date eventually anyway.
  - jdbernard 1y ago
    > but the second one doesn’t, the vowel is still there! Isn't the "a" in "have" elided along with the "h?" Shouldn't've Should not have What am I missing?
    - jack09268 1y ago
      Even though the vowel "a" is dropped from the spelling, if you actually say it out loud, you do pronounce a vowel sound when you get to that spot in the word, something like "shouldn'tuv", whereas the "o" in "not" is dropped from both the spelling and the pronounciation.
      - SAI_Peregrinus 1y ago
        The pronounced vowel is different than the 'a' in 'have'. And the "h" is definitely elided.
        int_19h 1y ago
        Many English dialects elide "h" at the beginning even when nothing is contracted. The pronounced vowel is different mostly because it's unstressed, and unstressed vowels in English generally centralize to schwa or nearly so.
  - dan353hehe 1y ago
    Don’t worry about us. English is truly a horrible language to learn, and I feel bad for anyone who has to learn it. Also I have always liked this humorous plan for spelling reform: https://guidetogrammar.org/grammar/twain.htm https://guidetogrammar.org/grammar/twain.htm
- theoreticalmal 1y ago
  My favorite variation of this is “oughtn’t to’ve”
amanaplanacanal 1y ago
That used to be the case, but "shouldn't of" is definitely becoming more popular, even if it seems wrong. Languages change before our eyes :)

YetAnotherNick 1y ago

Why not? Assuming you believe you can use any cloud for backup or Github for code storage.

solaire_oa 1y ago
IIUC one reason is that prompts and other data sent to 3rd party LLM hosts have the chance to be funneled to 4th party RLHF platforms, e.g. Sagemaker, Mechanical Turks, etc. So a random gig worker could be reading a .env file the intern uploaded.
- YetAnotherNick 1y ago
  What do you mean by chance? It's clear that if users have not opted out from training the models, it would be used. If they have opted out, it wont be used. And most of the users are in first bucket. Just because training on data is opt out doesn't mean business can't trust it. Not the best for user's privacy though.

gpm 1y ago

I think it's fair to question how proprietary your data is. Like there's the algorithm by which a hedge fund is doing algorithmic trading, they'd be insane to take the risk. Then there's the code for a video game, it's proprietary, but competitors don't benefit substantially from an illicit copy. You ship the compiled artifacts to everyone, so the logic isn't that secret. Copies of the similar source code have linked before with no significant effects.

FuckButtons 1y ago
AFAIK, the actual trading algorithms themselves aren’t usually that far from what you can find in a textbook, their efficacy is mostly dictated by market conditions and the performance characteristics of the implementation / system as a whole.
- short_sells_poo 1y ago
  This very much "depends". Many algo strategies are indeed programmatically simple (e.g. use some sort of moving average), but the parametrization and how it's used is the secret sauce and you don't want that information to leak. They might be tuned to exploit a certain market behavior, and you want to keep this secret since other people targeting this same behavior will make your edge go away. The edge can be something purely statistical or it can be a specific timing window that you found, etc. It's a bit like saying that a Formula 1 engine is not that far from what you'd find in a textbook. While it's true that it shares a lot of properties with a generic ICE, the edge comes from a lot of proprietary research that teams treat as secret and definitely don't want competitors to find out.
short_sells_poo 1y ago
Most (all?) hedge funds that use AI models explicitly run in-house. People do use commercial LLMs, but in cases where the LLMs are not run in-house, it's against the company policy to upload any proprietary information (and generally this is logged and policed). A lot of the use is fairly mundane and basically replaces junior analysts. E.g. it's digesting and summarizing the insane amounts of research that is produced. I could ask an intern to summarize the analysis on platinum prices over the last week, and it'll take them a day. Alternatively, I can feed in all the analysis that banks produce to an LLM and have it done immediately. The data fed in is not a trade secret really, and neither is the output. What I do with the results is where the interesting things happen.