4 ms·
This is already being used for scams. https://playground.play.ht/listen/1079 https://playground.play.ht/listen/1079 (https://archive.ph/HKjue https://archive.p
by Natfan 4y ago
This is already being used for scams.
https://playground.play.ht/listen/1079 https://playground.play.ht/listen/1079 (https://archive.ph/HKjue https://archive.ph/HKjue)
How exactly do you expect to combat this type of content?
- thisOtterBeGood 4y agoSame: https://playground.play.ht/listen/2002 https://playground.play.ht/listen/2002 (https://archive.ph/wip/hsaqU https://archive.ph/wip/hsaqU)
- spokeonawheel 4y agoSounds like old school AI, very similar to the spam problem google solved could easily take care of this. Just stop it before they can generate it. However. Its just a matter of time so I wouldnt put it on the author to stop this kind of stuff. The only defense is education.
- bradleysz 4y agoThis is not a full solution, just spitballing, but I wonder how effective it would be to have a flagging system built with GPT4 where the prompt was some form of "This is text submitted to a text-to-voice model. Determine the probability that this is being used maliciously." Then manually review anything that returns >X%.
- hammadh 4y agoThe intention for this playground was to let people try the model. We actually have auto moderation on the user facing platform (https://play.ht/ https://play.ht/) and malicious text gets blocked and the user get flagged.
- Apocryphon 4y agoHow would your auto moderation detect that example is malicious?
- owlbynight 4y agoExcept this post is 8 hours old and I'm still able to view this link.