to your point though, i can bet there are departments whose job is to apply data-fixes for these problems - since each prompt that stumps an AI is equivalent to a "bug"
the people who came up with this are obviously brilliant but being french myself, I really wonder why no one is proof-reading the english, this gives an overall bad impression of the work imho
Being a native English speaker I absolutely love reading and listening to speakers of English as a second language. Speaking is actually a subspecies of singing, and it's always cool to hear the same old lyrics remixed to a new melody and a new beat.
English has no 'correct' way to be written or spoken, nor does it need one, nor would it benefit from one, therefore, nor should it have one.
Speakers of English as a second language: you are what makes English a great language.
> English has no 'correct' way to be written or spoken, nor does it need one, nor would it benefit from one, therefore, nor should it have one.
There may be no 'correct' way, but there are plenty of 'incomprehensible' ways. I once encountered a research paper that had clearly [0] been translated word-for-word from French into English and made no sense until I translated it word-for-word back to French...
[0]: actually it was only clear after I realised I should attempt the reverse translation ;)
Sure, but frankly, I've heard plenty of people speaking the most flawless King's English who didn't make any sense at all.
re: translated math papers: haha we've all been there. Once I had to read a bunch of 70's-era papers from Russian Mathematicians. The translators, bless their hearts, I'm sure knew everything there was to know about Dickens and Dostoevsky, but it was clear they had no clue what the math was all about :-)
Oh well, Math is the universal language, right? chuckle
That's a beautiful way of seeing things! Unfortunately, as you're well aware I'm sure, most people do not share your idyllic view of polyglots and, for better or worse, they will assume that bad english = bad quality work. And bad doesn't have to mean mistakes. Just an unusual wording is enough to throw the average person off, in my experience.
I'm not as worried about those who have ears, but don't hear, as I am about the effect LLM's will have on English.
Grammerly was bad enough. One of my oldest friends is from Transylvania, and he could tell such great stories in his eastern-european accent and cadence. When he collected those stories into a book, he ran everything through grammerly, and the book reads like a soulless newscaster ;-(
When people start en mass to run their prose through LLM's to "correct" it, English will lose one of its main arteries.
Ha, I don't know anything at all about how language evolved. But, when you listen to somebody speaking--if you can bracket the meaning (which tends to soak up all our conscious attention)--you can hear the rhythm and you can hear the melodies. You can hear the music.
In order to understand somebody who speaks English in a different enough dialect, you have to really listen to the rhythm and melody--in order to puzzle out the meanings. The meanings are not hitting you in the face, they are more coy, and you have to seek them out while listening to songs you've never heard before!
Same goes with speaking with somebody who speaks English as a second language. You can hear the music in a way which is hard to do when listening to native speakers. Not impossible--once you realize what is happening, you can learn to pay attention to it.
But think about all the different ways you've hear English spoken...French accents, Nigerian accents, German accents, Russian accents, north Indian and south Indian accents, Mexican accents,.....It's like turning into a radio station playing the music of the world.
And unless they all were taking the time to learn English, we would not be hearing their music. And we would not be able to avail ourselves of an inexhaustible supply of new idioms, new ways of emphasizing, new ways of conveying subtle emotional cues...
if you're american and you're able to watch the video without feeling either ashamed or revolted, know that absolutely no one in the world respects the cloth you're cut from, except for the the worst dictators and their lackeys
I was thinking ABBYY FineReader, but those, too.
Instead of using VLMs or any sort of generative AI,
they're build on good old-fashioned feature extraction and nearest neighbor classifiers such as the k-nearest neighbors algorithm.
It's possible to build a working prototype of this technique using basic ML algorithms.
well done on the french achieving yet another extraordinary feat of engineering and research while still bearing the stigma of being shit at everything for some reason
I hope this international race ends up bearing fruit in a few decades, we need it
Zed has recently merged support for helix-like bindings, but it's still pretty far off (e.g. there's no way to default to "helix-normal" mode, meaning your first hotkey will be pure vim). I found it basically unusable for daily-driving, and Helix has me way too hooked on their style of modal editing for me to tolerate anything else.
I have been eyeing helix for quite a while. Been using (neo)vim for 22y. Why not Helix for the next 22!
I find the helix model of select then act so much more aligned to my mental model. Rather than verb then noun in vim.
Vim register handling on deletion is pure crap. (delete a line lands into a different register than delete a word, pasting over a selection overrides the default register etc.).
There is one plugin in vim that I cannot live without though, and that is easymotion (there are other similar plugins). It highlights objects with a letter (like the start of words for example), pressing the letter jumps to the object. I find this so much more natural and easier than jumping to line numbers. I might have to implement this into helix ;)
nah actually we should keep the slow, subpar tooling we got now because it came first
those people producing fantastic tools that have transformed the ecosystem of linting and packaging
should be working on it instead of working on transformative tech
