/jp/ - Otaku Culture

>>22209301
Is there any way to adjust timings for a group of sub files all in one go? With Aegisub, it seems you can only edit one sub file at a time.

Asking since I have 200 sub files which are all off by a whopping 43 seconds each.

Anyone know how to bulk-rename the subs from Kitsunekko in a way that accounts for the CRC codes in the filenames?

By CRC code, I mean the "[874C8EE7]" part in:
>[Coalgirls]_Hunter_X_Hunter_001_(1280x720_Blu-ray_FLAC)_[874C8EE7].mkv

I know how to rename
>[HorribleSubs] Hunter x Hunter - episodenumber [720p].srt
to
>[Coalgirls]_Hunter_X_Hunter_episodenumber_(1280x720_Blu-ray_FLAC).srt
en masse, but without the CRC code there to completely match the MKV's filename, my media player won't pick up the sub file.

https://ankiweb.net/shared/info/1213145732
Has anyone here managed to get this to work?

I get an error upon opening Anki after installing complaining about there being "No module named glob".

>>20863889
I'd much rather read and have a couple furigana mangas in my reading list, but I won't have the time before April. I figured I could set enough time for anki reps in the meantime because doing something meh is still better than doing nothing at all.

>>20801626
>>20801630
I'm going through Remembering the Kanji for reading, but what's a good audio source to start with other than anime? Pimsleur or is there something better?

>>20535852
You're not wrong to be skeptical, there is a chance the whole of research gets fucked overnight in a remake of the great AI winter, and it's true researchers are often a bit too optimistic (otherwise most would quit). I like this Bill Gates quote : We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.

Right now, the consensus is we lack something fundamental to get to General Artificial Intelligence (basically the very poorly defined thing you'd call intelligence). Some people smarter than me like Yann LeCun think it has to do with our poor understanding of unsupervised learning (the ability to infer interesting patterns from blobs of data without examples of a specific thing you're trying to find). Fortunately for most researchers "intelligence" is irrelevant as AI was a buzzword crafted to attract grant money even though it's really computational statistics by now (RIP symbolic AI).

Neural networks are shit at a LOT of tasks and in practice I avoid them as much as I can afford to for various reasons, but the architectures we have right now are also incredibly good at 2 things where they have driven extinct all competing approaches:
>Computer vision
>Natural language processing

In practice in most computer vision tasks you can think of (including and especially all of medical imaging), deep neural networks are now on par or superior to human expert performance. Having worked on vision research I can testify even a retard like me can fire up PyTorch and get something impressive done in an afternoon provided I have data, which is why humans have stopped being a relevant benchmark in research for a few years now.

Natural language is a bit tougher for now. Some tasks we've gotten fairly good at (sentiment analysis, image captioning) others are still very rough (text generation). Translation is somewhere in the middle, with performance depending on the type of text. People like to shit on Google Translate (which isn't even state of the art) but you can't deny it has improved significantly over the past few years, now you can translate a news article and get mostly intelligible text. I can't say for sure we'll have great translation bots in 20 years but I consider it a relatively safe bet considering what we know right now.

>>20535898
This war is over materialism has won. Sorry.

So Nazeka just got updated to have proper Anki support (automatic card creation + audio mining) in conjunction with the AnkiConnect add-on for Anki, but I'm not having any luck getting it to work on either 2.0 or 2.1.

Anyone else tried it with success?

>>18374319
>>18376881
I used to think that was just a joke. It's 14 years later and I'm not laughing any more.

>>16690200
Not as far as I know. There is Mozc which is based on the Google IME and that was actually installed on my system when I installed Japanese language support, but Mozc isn't showing up as an option there either...

>>12567228
K then.