I’ve used Descript a lot and it’s pretty handy for making small edits, but for anything longer than a short phrase, I still re-record my own real voice.
High quality AI voice cloning from short clips is very disturbing, and I'm trying to fight the urge to become paranoid. Have already warned my parents about the possibility to be on their guard. I definitely answer calls from unknown numbers less frequently now in part due to fear of giving out voice samples. I am also wary of branching into adding a podcast to my Substack, and wonder how high profile journalists and writers feel about this possibility? Do they take any precautions? Shrug it off as a risk of the job?
I think there's probably already enough audio out there that anyone who wanted to clone my voice could probably do it. But also I think all my close friends and family know about this technology now so hopefully they won't be fooled.
That’s a good point! Definitely seems like there’s an asymmetry to the risk: The *probability* of being cloned is higher for folks with a public profile, while the potential costs are higher for private citizens (people are more likely to fall for a scam of a close family member calling than Brad Pitt or Stephen King, and stand to lose a greater % of their $)
This is great, thanks Tim! I talked a little about upcoming US government regulation in my newsletter today; something like this is going to get a lot of attention.
I don't quite understand the need for this technology, and definitely all the dangers!
Yeah, it's not at all obvious the net impact will be positive.
Me too...
Super interesting.
I’ve used Descript a lot and it’s pretty handy for making small edits, but for anything longer than a short phrase, I still re-record my own real voice.
High quality AI voice cloning from short clips is very disturbing, and I'm trying to fight the urge to become paranoid. Have already warned my parents about the possibility to be on their guard. I definitely answer calls from unknown numbers less frequently now in part due to fear of giving out voice samples. I am also wary of branching into adding a podcast to my Substack, and wonder how high profile journalists and writers feel about this possibility? Do they take any precautions? Shrug it off as a risk of the job?
I think there's probably already enough audio out there that anyone who wanted to clone my voice could probably do it. But also I think all my close friends and family know about this technology now so hopefully they won't be fooled.
That’s a good point! Definitely seems like there’s an asymmetry to the risk: The *probability* of being cloned is higher for folks with a public profile, while the potential costs are higher for private citizens (people are more likely to fall for a scam of a close family member calling than Brad Pitt or Stephen King, and stand to lose a greater % of their $)
I’m starting to think that digital signatures / fingerprints will be necessary to establish content chains without generation.
I got a perfect score :D
Well done!
This is great, thanks Tim! I talked a little about upcoming US government regulation in my newsletter today; something like this is going to get a lot of attention.
https://trustedtech.substack.com/p/trusted-aicyber-news-and-notes-2023
Also, this has been floating around the public news sites today:
https://nypost.com/2023/04/12/ai-clones-teen-girls-voice-in-1m-kidnapping-scam/
Wow, I hadn't seen that story. Scary stuff.