There are a bunch of typos in there which jar a bit ('deterioted'), but I guess that makes sense for this specific article.
Personally, I would recommend them to simple use any old editor with spellchecking enabled. That suffices for most writing where you just want to keep your own voice. To me, the red crinkly line just means that I should edit that word myself. In the rare case where I'm stumped on the spelling I'll look at the suggested edit of course, but never as a matter of course.
The problem here, the overarching issue is that the subject complaint about AI slop is actually a bigger issue that has been plaguing America in particular for many years now, and of which the AI slop era is only a current top. The qualities of American writing have clearly been on a precipitous decline for a very long time now, predating AI slop and even spell checkers and computers.
Computers, digital text, and digital information distribution/transportation have made writing and thoughts cheap. Arguably due to what we are surely all aware of, humans rarely value that which is cheap, whether monetarily or in effort and consequential qualities. What people seem reluctant or maybe unable to acknowledge is that predating the current AI Slop, was what could be called Human Slop, low quality, low effort, careless output that was cheap; regardless of whether AI slop now outperforms.
It is why you are justified in pointing out that even in the post complaining about AI Slop, the human has apparently abandoned what would have been common practice in just the recent past, using basic spellcheckers or simply reviewing what was written and also practicing with deliberation; the art and skill of writing, grammar, and sentence structure.
No one is perfect and that is also what makes anything human, somewhat inexplicable and random variation. However, it takes a certain refinement before unique human character becomes a positive quality and is not just humans being sloppy ... human slop.
> The qualities of American writing have clearly been on a precipitous decline for a very long time now, predating AI slop and even spell checkers and computers.
> Every NYT bestseller from 1960 to 2014 falls in the seventh-grade level spread, from 4th to 11th.
> ...
> Since 2000, only 2 bestsellers have scored higher than 9th-grade readability.
> ... ...
> The bestselling authors of our time are writing at the 4th-grade level.
> > “8 books tie for the lowest score,” a 4.4, just above 4th-grade level. Prolific, well-known authors with huge sales: James Patterson, Janet Evonvich, and Nora Roberts.”
> These three authors have written a combined total of 419 books.
Whenever I read something from roughly the first half of the 20th century (I'm not sure where the cutoff point is, it seems to the 1960s), I'm struck by the quality of the writing. I'm not sure what happened, but it's pretty clear that at some point we stopped taking ourselves seriously.
We see the same thing in how people dress. People used to write "respectably", and they used to dress the same, and in TV interviews they spoke with great care and deliberation.
> They could released a Chrome extension to let users configure their links for each of those apps. Wasted effort.
This app isn't just some link aggregator or an admin dashboard, though. It's workplace software that hosts all your data, self-hosted on your system of choice if you wish. I'm neither a user of nor am I affiliated with this project, but it seems like there's the aspiration to provide a unified client interface for every app, and it looks like you could BYOC as well (for CalDAV and Email).
Just tried it with B.E.D - Walk Away[0], unfortunately it lost track of the lyrics after 30 secs (Model is "large-v3"). Will play around a bit more, as it would be great to have a working karaoke generator.
Some quick feedback:
- Needs a way to skip for-/backwards during playback to validate the result
- Sentences seem to be recognized (first letter has uppercasing), but periods aren't added
- Needs an option to edit results from the track analysis
indeed, I'm running to two problems on the analyzer side:
1. align model sliding off (especially w/ chorus/back vocals present)
2. transcript skipping parts of lyrics in lyrics-heavy tracks (I tried a lot of russian rap, lol)
happy for contributions as I'm not that experienced w/ machine learning side of the project, mostly it was emperical "tweak the parameters and look what is changed"
Bought a used MacBook Air M2 past summer to run Asahi linux exclusively on it, the installation went hassle-free. One charge lasts 9+ hours easily, sometimes up to 12 hours. Thunderbolt, DP Alt Mode and TouchID would be nice to haves, but I'm super happy how everything runs. Thank you everyone on the Asahi team!
I think the support for linux/arm64 is already very good in general, can't answer on pytorch though. The only app I'm really missing is Signal Desktop. The virtualization to run games is a noticeable performance hit and shows occasional glitches in the Steam overlay, but all my games run smoothly.
I bought my last Acer around 2010 (Aspire 4820TG I think, good machine). Their notebooks were always on the cheaper side, where its price just sat right with the offered value. Cooling issues were always present and weren't a big problem as long as the machine was maintainable. Unfortunately maintainability in notebooks (and electronics in general) all changed around 2015-ish and from there on it was used ThinkPads only for me.
Do you care to elaborate? "if (...) return ...;" looks closer to an expression for me:
reply