5 ms·

It's only a matter of time before we'll see auto-generated content that uses AI-generated imagery, text and voice synthesis to produce content that appears vagu

It's only a matter of time before we'll see auto-generated content that uses AI-generated imagery, text and voice synthesis to produce content that appears vaguely intelligble enough to fool people to watch it. And as the downvote count is private information now it will be very hard to distinguish these videos from legitimate ones. Personally I think creative-commons licensed content is a double-edged sword. It seems every month there's a new Stackoverflow clone popping up, which just blatantly copies all content from there, remixes it a bit and spams the Google index with it. And surprisingly, Google seems unable or unwilling to filter this stuff out. Videos of some of my conference talks were also published under a creative commons license (without my explicit consent) as the organizers wanted to make them available to the world for free, which in principle is a good thing. This only led to people blatantly reposting them on Youtube and misattributing them to farm views. I no longer not agree to any of my talks being released under such a license therefore.

charcircuit 4y ago
If it's a good video why should people care. If my favorite content creator turned out to be an AI that wouldn't suddenly cause me to think that their content is now garbage.
philonoist 4y ago
Please make your dislike count visible to you like us. 1. https://chrome.google.com/webstore/detail/return-youtube-dislike/gebbhagfogifgggkldgodflihgfeippi https://chrome.google.com/webstore/detail/return-youtube-dis... So far as your downvote is concerned, should there be crowdsourced information of a video made out of AI on it? I could get inspired from - https://chrome.google.com/webstore/detail/sponsorblock-for-youtube/mnjggcdmjocbbbhaepdhchncahnbgone https://chrome.google.com/webstore/detail/sponsorblock-for-y... The video will have a small "%/absolute numbers think this is an AI fused video." beneath it. I think we can wait on this until a more robust solution comes along.
netsharc 4y ago
It feels like there are already audiences who are already fooled by content narrated by robots, I've seen some where they don't sound totally robotic, they just mispronounce particular terms that you know it's speech synthesis...
- Akronymus 4y ago
  I keep getting those reddit story reads recommended to me, because I accidentally clicked on one ages ago. Just pure text to speech, hell, probably completely automated (scrub top of week of x subreddit, then make the TTS read it out)
bambax 4y ago
> And surprisingly, Google seems unable or unwilling to filter this stuff out. A little OT, but I find this very surprising too. In general I'm (still) a big fan of Google Search, which I think is amazing and miles above the competition. But it's true that SO clones are everywhere, often in the first five results. How can this be? How hard is it to detect that if the exact same content appeared first on SO, and then later on ten different sites that have zero content that's not already on SO, those are spam clones?? How is it even in Google's interest to upset users that way? Beats me.
- londons_explore 4y ago
  I suspect the problem is that Google can't fully crawl stackoverflow/github etc. There are plenty of pages which appear unindexed. I find it a lot when doing a google search for a code snippet and getting no results, only to find that exact piece of code/text is in a public github repo or bugreport. Then, the clones copy those pages, and do manage to get indexed. Which in turn means that when you search for something, the clone appears to be first. I think Google just needs to make their indexer 10x more aggressive, especially for big popular sites.
  - quickthrower2 4y ago
    They probably need to hard code in some trust levels of sites. I suspect they want everything to be algorithmic through and not involve human judgment. Same content on SO and other site means always prefer SO even if it was indexed second
    - bambax 4y ago
      Yeah that's probably a factor. They don't want to go the route of manual intervention. But they should have other sources of information that they can use.
  - akudha 4y ago
    I suspect the problem is that Google can't fully crawl stackoverflow/github etc. How is this possible, given their insane resources and expertise? If they can't crawl fast enough, can't they simply make a deal with Microsoft to get the public data from Github the moment it is created? I assume they have similar deals with other companies (Twitter maybe? for their firehose?). This particular thing doesn't seem to be a tech limitation. Most likely Google doesn't care enough to fix the issue, they could probably fix it in a week if they wanted to. At this point, nothing is going to improve until a serious competitor to Google comes along. And I don't see that happening anytime soon, so all the issues that users have with Google are likely going to stay.
Foobar8568 4y ago
It has been done for at least a decade, back when the scandal poped out back in 2017. https://www.avclub.com/take-a-trip-to-the-automated-hellscape-of-youtube-video-1820196139 https://www.avclub.com/take-a-trip-to-the-automated-hellscap... And in a more general way https://en.wikipedia.org/wiki/Elsagate https://en.wikipedia.org/wiki/Elsagate
jobigoud 4y ago
> AI-generated imagery, text and voice synthesis to produce content that appears vaguely intelligble enough to fool people to watch it You seem to readily assume this content isn't worth watching. People consume all sorts of content for all sorts of reason.
aeyes 4y ago
> It's only a matter of time before we'll see auto-generated content that uses AI-generated imagery, text and voice synthesis to produce content that appears vaguely intelligble enough to fool people to watch it. This is already happening, look at the following of Hololive. For now there are humans behind the animated figures but they aren't much more than story writers and it shouldn't take long for the company to figure out that they could automatically generate this content.

charcircuit 4y ago

If it's a good video why should people care. If my favorite content creator turned out to be an AI that wouldn't suddenly cause me to think that their content is now garbage.

philonoist 4y ago

Please make your dislike count visible to you like us. 1. https://chrome.google.com/webstore/detail/return-youtube-dislike/gebbhagfogifgggkldgodflihgfeippi https://chrome.google.com/webstore/detail/return-youtube-dis... So far as your downvote is concerned, should there be crowdsourced information of a video made out of AI on it? I could get inspired from - https://chrome.google.com/webstore/detail/sponsorblock-for-youtube/mnjggcdmjocbbbhaepdhchncahnbgone https://chrome.google.com/webstore/detail/sponsorblock-for-y... The video will have a small "%/absolute numbers think this is an AI fused video." beneath it. I think we can wait on this until a more robust solution comes along.

netsharc 4y ago

It feels like there are already audiences who are already fooled by content narrated by robots, I've seen some where they don't sound totally robotic, they just mispronounce particular terms that you know it's speech synthesis...

Akronymus 4y ago
I keep getting those reddit story reads recommended to me, because I accidentally clicked on one ages ago. Just pure text to speech, hell, probably completely automated (scrub top of week of x subreddit, then make the TTS read it out)

bambax 4y ago

> And surprisingly, Google seems unable or unwilling to filter this stuff out. A little OT, but I find this very surprising too. In general I'm (still) a big fan of Google Search, which I think is amazing and miles above the competition. But it's true that SO clones are everywhere, often in the first five results. How can this be? How hard is it to detect that if the exact same content appeared first on SO, and then later on ten different sites that have zero content that's not already on SO, those are spam clones?? How is it even in Google's interest to upset users that way? Beats me.

londons_explore 4y ago
I suspect the problem is that Google can't fully crawl stackoverflow/github etc. There are plenty of pages which appear unindexed. I find it a lot when doing a google search for a code snippet and getting no results, only to find that exact piece of code/text is in a public github repo or bugreport. Then, the clones copy those pages, and do manage to get indexed. Which in turn means that when you search for something, the clone appears to be first. I think Google just needs to make their indexer 10x more aggressive, especially for big popular sites.
- quickthrower2 4y ago
  They probably need to hard code in some trust levels of sites. I suspect they want everything to be algorithmic through and not involve human judgment. Same content on SO and other site means always prefer SO even if it was indexed second
  - bambax 4y ago
    Yeah that's probably a factor. They don't want to go the route of manual intervention. But they should have other sources of information that they can use.
- akudha 4y ago
  I suspect the problem is that Google can't fully crawl stackoverflow/github etc. How is this possible, given their insane resources and expertise? If they can't crawl fast enough, can't they simply make a deal with Microsoft to get the public data from Github the moment it is created? I assume they have similar deals with other companies (Twitter maybe? for their firehose?). This particular thing doesn't seem to be a tech limitation. Most likely Google doesn't care enough to fix the issue, they could probably fix it in a week if they wanted to. At this point, nothing is going to improve until a serious competitor to Google comes along. And I don't see that happening anytime soon, so all the issues that users have with Google are likely going to stay.

Foobar8568 4y ago

It has been done for at least a decade, back when the scandal poped out back in 2017. https://www.avclub.com/take-a-trip-to-the-automated-hellscape-of-youtube-video-1820196139 https://www.avclub.com/take-a-trip-to-the-automated-hellscap... And in a more general way https://en.wikipedia.org/wiki/Elsagate https://en.wikipedia.org/wiki/Elsagate

jobigoud 4y ago

> AI-generated imagery, text and voice synthesis to produce content that appears vaguely intelligble enough to fool people to watch it You seem to readily assume this content isn't worth watching. People consume all sorts of content for all sorts of reason.

aeyes 4y ago

> It's only a matter of time before we'll see auto-generated content that uses AI-generated imagery, text and voice synthesis to produce content that appears vaguely intelligble enough to fool people to watch it. This is already happening, look at the following of Hololive. For now there are humans behind the animated figures but they aren't much more than story writers and it shouldn't take long for the company to figure out that they could automatically generate this content.