Microsoft’s new AI can simulate anyone’s voice with 3 seconds of audio

Text-to-speech model can preserve speaker's emotional tone and acoustic environment.

Ars Technica
@drewharwell So happy they're using AI to solve important problems like 'impersonate voices' or 'put writers and artists out of work', instead of crap like 'fix climate change' or 'find an equitable solution to institutionalized racism'.
@el_rubino @drewharwell You see, the latter would involve the powerful making sacrifices, and that won't be allowed, so instead they go after powerless people such as artists. Just disgusting.
@el_rubino @drewharwell There is such an organization - https://aiformankind.org/ - but of course it's a nonprofit and the work is being done by volunteers.
AI For Mankind

Our volunteers build AI+Data solutions to solve world challenging problems. Together we can make things happen !

AI For Mankind
@el_rubino @drewharwell I mean, that's the thing. "AI" isn't actually *intelligent* - it isn't going to find solutions to systemic social problems.
You wouldn't ask your toaster to prepare a meal, etc.
@el_rubino @drewharwell No training data exists for those latter cases 🙃

@el_rubino @[email protected] Why else would they build AI if not to grow their power base? To do otherwise would be foolish.

They may not be wise or helpful, but they do know what side their bread is buttered on.

@drewharwell "My voice is my passport. Please verify me."
@ajhekman @drewharwell not many will get that, but I did 🙂
@drewharwell Angelina Jolie just called me and said she wants to meet me for dinner! So I sent her money for Uber
@drewharwell There go all the local radio guys who do voiceovers for their respective stations.
@theloneapple @drewharwell Local radio guys? I thought they were already gone.
@Klaxun @drewharwell There are the underpaid ones who have to do all the production.
@drewharwell minor inconvenience in comparison what AI will do to us next 😅🦾
@drewharwell Please note that this call may be recorded for “training” purposes. (Training. Nyuck Nyuck)
@drewharwell I will combat this by wildly varying my pitch and intonation at all times
@drewharwell why I don’t pick up the phone anymore to numbers I don’t recognize. Leave a message and if it is legit, I will call back and/or plug the number into my list.
@drewharwell so recordings and images are now *completely* untrustworthy - will society learn to discount recorded media as sources of truth?
@drewharwell I was thinking today that scammers are going to have a field day with this technology 
@drewharwell you mean like when we get spoof robocalls that get people to say things like “Hello? Who is this?” or “Sorry, not interested, thanks.” Boom, mass collection of 3s snippets tied to a particular phone number.
@drewharwell I don't see the problem here. If you don't want your voice being turned into an AI that can make it sound like you're saying a bunch of stuff you never said, just don't speak 🤷

@willoremus @drewharwell

no no you don't understand there's an ethics statement

"The experiments in this work were carried out under the assumption that the user of the model is the target speaker and has been approved by the speaker. However, when the model is generalized to unseen speakers, relevant components should be accompanied by speech editing models, including the protocol to ensure that the speaker agrees to execute the modification and the system to detect the edited speech."

@willoremus Every six weeks for security I change my voice, face, eyeballs, gait, writing style, license plate and palm veins
@drewharwell wise. imo it's a small price to pay for.... whatever it is we're supposed to be getting in return
@willoremus I can buy chicken nuggets online two seconds faster

@drewharwell @willoremus Yep, but changing smell, in order to outsmart #olfactory #AI , will be quite... difficult:

https://aryballe.com

https://www.forbes.com/sites/bernardmarr/2021/05/10/artificial-intelligence-is-developing-a-sense-of-smell-what-could-a-digital-nose-mean-in-practice

Anyway, on the plus side, "Olfactory AI" (pronounced "𝘖͛𝘭͛𝘧͛𝘢͛𝘤͛𝘵͛𝘰͛𝘳͛𝘺͛ ͛𝘖͛𝘠͛") would be a good name for a Finnish #rock band ✌🤘

Aryballe | SIMPLE ODOR ANALYTICS

@drewharwell robots.txt, but in my mouth
@[email protected] for the reassurance. My voice is well and truly out there. I'm doomed.
@drewharwell Mit irgendwelchen Fake-Anrufen geht das schon. Da reicht den aktuellen Betrügern ja schon ein JA um einem einen Strick draus zu drehen,
@drewharwell @martinvermeer You mean like a voicemail that you leave?
@drewharwell Disturbing. Luckily, it works well only if you read a book out loud and donated the recording to #LibriVox.
@drewharwell “Alexa, mhhmmm” (swallowing hard)
Good thing we don't send each other voice messages over every platform or run around with portable microphones and cameras everywhere and spend our time recording ourselves voluntarily only to upload hours of footage to servers of companies who have a bad track record of using that data in a privacy respecting way.

@drewharwell "it is possible to build a detection model to discriminate whether an audio clip was synthesized by VALL-E."

I wish they had said clearly that they were going to build that and require it stay detectable. All the banks wish that too.