Thanks! I’m self-hosting it, and it’s currently just me using it. I had a few spare VPS systems and figured I’d try running Lemmy and Mastodon on one of them.

Traditional OCR isn’t AI; it relies on manually-written rules. Some modern OCR tools use AI concepts (e.g. Tesseract uses a neural network) but they don’t necessarily have to. Getting humans to manually enter words is definitely not AI.

They have a LOT of compute power… They could have several baked in ads per geographical area / demographic and only store them on servers in / close to the relevant country. There’s definitely associated costs but I wonder if it’d amortize well given their viewer count.

PGP is a different piece of software though. Would you refer to Firefox as “Chrome” because both of them can use the same protocol (HTTP)?

This reminds me of my parents referring to every games console as a “PlayStation” lol

If the ad is just blacked over, it is still loaded, and they wouldn’t know.

If it became a common thing, they’d have to add detection for it. Not necessarily to stop people doing it, but to ensure advertisers aren’t charged for invalid impressions. Practically every major ad network has adblocking detection; they just don’t always make it obvious (e.g. they might silently log it).

Yes! 80% of home ownership is essentially water management.

Stop using giant catchall instances and switch to a smaller instance that’s more suited to you.

One of the major advantages of a federated system is that it doesn’t really matter which instance you use. There’s no real advantage to using a larger instance, and in fact there’s several disadvantages as the large instances can be slower, maintenance can take longer, it’s more expensive to run the servers, etc.

One of the reasons people moved away from Reddit was to avoid one company (Reddit) and especially one person (the Reddit CEO) having control over the whole thing. Using a huge Lemmy server kinda defeats the point of switching across.

I prefer using the “scaled” feed rather than “active”. It’s like active, but boosts posts from smaller communities, and seems to usually surface newer content.

Please don’t.

Some people like mbin. Some people like Lemmy. Some people like Kbin. For the most part, they can communicate with other. Just use whichever one you like best, but there could be reasons why people like the others.

then it would process the file and remove the commercials

This still exists today, for example in Plex’s DVR. Practically everything that blocks commercials these days uses comskip or a fork of it.

I paid for it before they removed Google Play Music. I was on one of the plans that was $8 for both Google Play Music and YouTube Red.

YouTube is forced to display the “Advertisement” mark

They’re forced to identify that it’s an ad, but they don’t have to do it in a machine-readable way. There’s many different approaches to show an “Advertisement” or “Sponsored” label that appears to users but that blockers can’t easily find.

Let’s go full guerilla: Plugin that lets you select the first and the last frame of an ad, thus allows to report the beginning and length to a synced database. When that frame is found in the buffer, skip X frames ahead.

This would fit in well with SponsorBlock, which already does the same thing for different parts of videos (eg sponsored segments, intro and outro animations, non music segments in music videos, etc).

I suspect YouTube will find ways around this, like running ads of differing lengths, add random amounts of padding at the start of the video or between ads, etc.

I don’t care if YouTube gets paid for it

Legally, YouTube have to detect if ads were blocked and and mark the impression as non-billable. They can’t charge advertisers for blocked ads.

No it wasn’t… It was human-assisted OCR to help digitize books. Initially for Project Gutenberg, but then for Google Books once Google acquired it in 2009.

Then Google acquired them. Text CAPTCHAs got phased out

Google kept the text version for five years after the acquisition though. They used it to digitize books on Google Books, to allow full-text search of their book archive.

The original reCAPTCHA from Carnegie Mellon University was helping to digitize books. It showed one known word and one unknown word, and if enough people answered the second word with the same answer, that’d be marked as the correct value.

Aren’t they going to have the same issue though?

lol at the DO NOT TRUST keys.

I’ve learnt over the years that you have to make the example code fail to compile or print out huge user-visible warnings or something like that, otherwise people can and will use it as-is in production, hard-coded keys and all.

Even if you make it print out a huge message, some manufacturers will just comment that out while keeping all the other dummy example data.

I’ve seen several production OAuth/OpenID servers that accepted an app ID and secret from a “how to set up an OAuth server” tutorial, and in one case the company was using that app ID for all their production services.

Right. OpenPGP is the protocol. PGP is the original app, which predates the spec.