from Hacker News

Will AI generated speech get banned from social media/YouTube?

by dfps on 12/21/22, 5:56 PM with 12 comments

• by valdiorn on 12/21/22, 8:49 PM

  There's millions upon millions of "top 5 list of X" videos on YouTube already, with auto-generated synthesized voice and some low quality slideshow with screenshots grabbed off the internet.

  If it gets clicks, YouTube won't ban it.

• by logicalmonster on 12/21/22, 6:15 PM

  AI generated creative work has the capacity to be great and highly entertaining.

  Where I think some people are going to screw themselves (and possibly the entire industry) is that somebody is going to get very greedy and just mass produce a bunch of nonsensical low-quality videos that might get some views through good SEO, but that nobody really loves.

  I can't imagine that YouTube really wants to deal with hosting multiple Terrabytes or more of low-quality content from one random person that just puts no thought or effort into the videos and is essentially creating a content farm that costs Google a bunch of resources to deal with.

• by PaulHoule on 12/21/22, 6:16 PM

  There are numerous videos on YouTube with AI voices already, in fact you hear the same AI voices a lot on ads for scam products like supplements. For a while there were a number of (I think) Ukrainian propaganda videos made with voices that don't know how to pronounce Ukrainian words... (e.g. "Kherson" doesn't start with a "K" sound)

  Personally I'd see it as a game changer if text-to-speech plus a little direction could get a result closer to hiring a voice actor.

• by arduinomancer on 12/22/22, 3:20 AM

  I've seen a few discussions about this on HN but I don't understand, why does everyone assume its even possible to ban it?

  How would that even work without false positives?

  For example with comments, in the worst case an AI model could just output a real comment from the training set verbatim

• by muzani on 12/22/22, 3:09 AM

  "No one likes the content"

  This is really the core problem, not AI. The Play Store is already full of lucrative garbage. But people download and pay, so by all metrics it's fine.

  There's lot of artificial speech in memes. VTubers have been around for a long time and are very popular. Vocaloids too. If someone wants to ban "AI" they will have to define it separately from low quality autogenerated content, and it's easier to just ban low quality content.

• by btbuildem on 12/21/22, 8:40 PM

  I don't think so. There's already a lot of this type of content (I'm sure you're familiar with "that tiktok voice") -- people hate it, but I guess overwhelmingly more people don't mind it.

  Social media in general is a cesspool of mediocrity, so this type of thing fits right in.

• by tluyben2 on 12/21/22, 7:13 PM

  I cannot see much issues with realistic generated speech if the content is not AI written and high quality. There are plenty of people who are not very good at speaking but good at writing or being able to create interesting content.

  Low quality posted by spambots I would not mis.

• by shanebellone on 12/21/22, 6:13 PM

  I'd ban it now if any of those platforms were mine. However, they have different ways to deal with it. Content, as described, will not rank. With that in mind, I doubt this problem will actualize.

• by yellow_lead on 12/21/22, 6:05 PM

  There was a post of a YouTube account that posts auto-read text every ~5 seconds or so and they have thousands of views.

• by Beaver117 on 12/21/22, 8:19 PM

  There are probably billions of garbage low effort clickbait videos using text to speech. Why would they care now?

• by ipaddr on 12/21/22, 11:45 PM

  It's already against the terms.