Current Events > What do you think of AI copying people's voices?

(+) Yeah! (+)

At least some of the services require under a minute of audio to copy your voice, and with the exception of certain accents, the results are perfectly convincing.

That "recording" of you confessing to a crime a prosecutor presents to the court?
The "secret recording" your ex-wife presents in a custody hearing where you threaten to hurt your children?
That "conversation" your boss or coworker recorded where you admitted to unethical or illegal behavior?
The audio "recording" that the spam caller has of you enthusiastically agreeing to be billed for services?
That call your elderly parents received where you begged them for money or personal information, with a number that looked just like yours (phone number spoofing is a thing, if you didn't know)?

Yeah, none of those have to really be you, and in the short to medium future, I don't know how society is going to come to grips with this as knowledge of the technology and its use spreads. It will also impact all types of professional voice work, and make misinformation much easier to propagate. Embarrassing or confusing or even criminal deep fake photos are well on there way (but not great yet), and video will follow in likely just a few years.

I've never heard a robot voice sound natural in Norwegian.
Guess we're in the clear for the time being.

The tech is moving way faster than the law in this area. Even some AI developers have asked for a global pause. Six months, IIRC.

brestugo posted...

Even some AI developers have asked for a global pause. Six months, IIRC.

they don't really have a way to enforce that, except among the major platforms. anyone can train and run GANs and LLMs on their local machines, and undetected at that.

Worrisome

Cocytus posted...

Worrisome

It is, to quote one of the heavy thinkers of our time; concerning.

If we get a Samuel L. Jackson GPS, it's all worth it.
"Turn left in a quarter-mile, motherfucker!"

TheMikh posted...

they don't really have a way to enforce that, except among the major platforms. anyone can train and run GANs and LLMs on their local machines, and undetected at that.

No, but it shows there's a level of concern within the industry.

https://m.youtube.com/shorts/cIQ4lf6vkJo

Fun for goofing around but there definitely needs to be hard legislation in place.

TheMikh posted...

anyone can train and run GANs and LLMs on their local machines, and undetected at that.

For now. It's not inconceivable that laws might eventually be passed to require makers of operating systems to periodically scan your devices for unauthorized AI technology (governments would already like to do this for child abuse material, drugs, and terrorism related content). An easier first step to control might be making the unauthorized distribution of AI materials illegal. If vigorously pursued, that would eventually lead to AI technology being pushed to the dark web and reduce access.

I in no way advocate for any such measures, but it is unrealistic to think they might not be tried. I'm not particularly optimistic that misuse by governments, police, and prosecutors will be addressed with any speed (or ever). I think there will be a reckoning sooner or later regarding society's inability to address the problems with inequality, policing, and the justice system.

There is undoubtedly some good use to the technology, but it's moving extremely quickly and the bad use cases are a lot more prevalent and harmful.

At the end of the day the key issue is ownership of one's self. Your voice is your voice we need strict legislation that protects your ownership of your voice with strict punishments for those who try to use it without your permission. This isn't to say that AI voice technology should cease. But that it needs to be paying for the voices it uses and that clearly defined contracts should be in place over their use.

Dangerous.

The tech industry has been living out that Jurassic Park quote for decades.

I think we need laws on this asap.

Make crimes using AI get harsher punishments for now too.

I think it's cool, especially for mods that now can be almost like the real thing. Also like the weird YouTube mashups that have Sinatra singing ridiculous stuff. This definitely should fall under parody or fair use as long as they don't profit from them.

What's not ok is companies possibly using this as a means to pay less or avoid using voice actors. They really need to renegotiate so they can legally use an actors looks or voice beyond the original scope of the project.

xlr_big-coop posted...

What's not ok is companies possibly using this as a means to pay less or avoid using voice actors. They really need to renegotiate so they can legally use an actors looks or voice beyond the original scope of the project.

This is a big issue for sure. AI has the potential to kill voice acting as a profession. I really hope we don't let that happen.

itachi15243 posted...

I think we need laws on this asap.

Make crimes using AI get harsher punishments for now too.

I've heard a story or two about AI being used to spoof someone's voice as part of a "ransom scam" (for lack of a better term). Not sure if that's an urban legend or not though.

brestugo posted...

I've heard a story or two about AI being used to spoof someone's voice as part of a "ransom scam" (for lack of a better term). Not sure if that's an urban legend or not though.

Many criminals use this tecnology for this purpose, and ever you can fake a voice through AI. I think AI should be legislated and it needs to have a proof to know this voice, picture or other AI related applications were made with AI tecnology.

A bad idea

darkmaian23 posted...

It's not inconceivable that laws might eventually be passed to require makers of operating systems to periodically scan your devices for unauthorized AI technology (governments would already like to do this for child abuse material, drugs, and terrorism related content).

i'd imagine microsoft already has backdoors. apple has demonstrated a reluctance to infringe on customer privacy though their customers' cloud services seem to get hacked like it's going out of style anyway, and they could probably be pressured into compliance by virtue of being a public company anyway.

at the end of the day though, like ai, it's virtually impossible to regulate every operating system developer - a lot of more sophisticated deep neural networks are being deployed and trained on open-source unix-like systems which are inherently resistant to state intervention. authorities would have to target the hardware manufacturers themselves, which would cause a lot of otherwise disinterested people to sound alarms.

An easier first step to control might be making the unauthorized distribution of AI materials illegal. If vigorously pursued, that would eventually lead to AI technology being pushed to the dark web and reduce access.

this is far easier said than done. far more likely to be pushed to the deep web where it can be torrented and such, rather than the dark web proper, and it would be a setback to positive independent innovation that would undermine national industries who take that approach down the road.

I in no way advocate for any such measures, but it is unrealistic to think they might not be tried. I'm not particularly optimistic that misuse by governments, police, and prosecutors will be addressed with any speed (or ever). I think there will be a reckoning sooner or later regarding society's inability to address the problems with inequality, policing, and the justice system.

that's reasonable.