PCG Article Hear AI Todd Howard declare The Elder Scrolls 6 will let players have sex with NPCs

[ZedClampet]

An older story that I missed until this morning. I didn't put it under RPG because I think the discussion is mostly about ethical uses of AI and using this tech in game development.

First of all, if you have an immature mind like I do, you should listen to the video of AI Howard talking because you'll probably find it hilarious. Be careful where and how you listen to it, though. It is NSFW.

What strikes me most, though, is how you would never know it wasn't Howard talking. It sounds perfectly real to me except, of course, for the content. Per the article, the tech is currently being used to create mods that give NPCs more lines. You just train the AI on the existing NPC lines, and then the AI can say whatever you want it to.

Now, I haven't heard any other instances where this was used, but the AI "acting" in the Howard piece was very convincing, but that was supposed to be a talk he gave and didn't require a lot of emotion. Of course, any time we qualify AI with something like that, it's really only a "matter of time" rather than "AI can't do it". But I might see if some of these mods have preview videos and post them later in this thread so we can see what they sound like.

Of course, this AI comes with a lot of unsavory possibilities, like uncredited, unpaid use of someone else's voice and harmful fakes. Between this and face swapping AI, I think there really needs to be quick legislation that controls all of this and puts standards in place. States like California have already got laws on the books about using unsuspecting people in adult material, but I think it needs to go further.

Your opinions on all of this?

Oh, and unbeknownst to anyone, Audacity is capable, somehow, of doing this at least to some small degree. I've had two occasions where Audacity has changed what someone on a recording I made said. It shocked me at the time. Now I understand what was going on. Audacity is open source. I'm not sure it's supposed to be doing that. Probably it's a prank someone coded into it.

 

[Frag Maniac]

I'm not sure it's supposed to be doing that. Probably it's a prank someone coded into it.

Uh, yeah Audacity has never had that capability. The most I've ever done with it is change the pitch of the audio a bit, but usually just the volume of background music I add to some of my videos. Since YouTube went ballistic editing licensed audio out of video clips (by cutting out the video's audio holistically), which is something that often doesn't happen for a year after you upload it due to copyrights of the content often changing hands, it's been a long time since I've used background music.

As for that "video", I'm not sure it's even safe for public forums that allow a broad range of ages to participate, and I get the feeling those whom use that software that way are eventually going to be subject to prosecution.

That said, sometimes when I've spotted weird things NPCs do in games, I get a little creative dubbing in game dialog to go with...

 

[Zloth]

Don't forget the reverse implications of this: "I never said that! They just used AI to make it sound like I did!"

States like California have already got laws on the books about using unsuspecting people in adult material, but I think it needs to go further.

Will that help any?? Even if that was a national law, all you need is one nation on the internet that doesn't have a similar law and it could serve up videos galore.

Anyone remember the Babylon 5 TV-movie "River of Souls"? That had a parlor where a customer would bring in a few old photos of somebody, put on a tactile suit that could simulate touch, and a hologram of the 'somebody' would <ahem> service them. The photos could be a famous person or just the old college refused-to-be-your-girlfriend, and the AI would do the rest. No holograms for our situation, but it's the same idea. +1 to Straczynski for seeing that one coming.

 

[mainer]

An older story that I missed until this morning. I didn't put it under RPG because I think the discussion is mostly about ethical uses of AI and using this tech in game development.

First of all, if you have an immature mind like I do, you should listen to the video of AI Howard talking because you'll probably find it hilarious. Be careful where and how you listen to it, though. It is NSFW.

I mentioned that article in an earlier post this week titled: "Three recent PCG articles on sex & romance in games", or something to that affect. But I also found it hilarious, so I guess I have a bit of an immature streak myself.

Per the article, the tech is currently being used to create mods that give NPCs more lines. You just train the AI on the existing NPC lines, and then the AI can say whatever you want it to.

Now, I haven't heard any other instances where this was used, but the AI "acting" in the Howard piece was very convincing, but that was supposed to be a talk he gave and didn't require a lot of emotion. Of course, any time we qualify AI with something like that, it's really only a "matter of time" rather than "AI can't do it". But I might see if some of these mods have preview videos and post them later in this thread so we can see what they sound like.

Of course, this AI comes with a lot of unsavory possibilities, like uncredited, unpaid use of someone else's voice and harmful fakes. Between this and face swapping AI, I think there really needs to be quick legislation that controls all of this and puts standards in place. States like California have already got laws on the books about using unsuspecting people in adult material, but I think it needs to go further.

I have a great interest in this kind of AI tech and how it could eventually be used in games, especially for NPCs, companions, and other speaking characters. To me it relates back to a post I made in your thread about, "a game you'd like to make" (Sorry, I can't remember the title of that thread). I mentioned about wanting to create a reactive AI for characters to react to you and world events. It's a leap of logic on my part, but I think that the AI vocal synthesis used in that video could be the first step towards developing that reactive AI I dream about.

As these various AI programs continue to develop, and they will, we're going to have to have new laws or controls on how they are used. There's going to be a lot of arguments about what constitutes "fair use" and "copyright infringement". Can a professional voice actor's voice be copyrighted? Their voice is their livelihood, so they should have a say if, or how, it's going to be used, as well as receive adequate monetary compensation.

If I was a game studio, and I wanted to hire Jennifer Hale (who I think is awesome) for the voice of one of my characters, I'd have her record a certain amount of dialogue in my sound studio that the AI program could learn from. She would be fully compensated financially, made aware of how the AI was going to create more lines of dialogue as it learned her voice patterns, fully credited for her work, yet wouldn't be required to record every single line spoken within the game.

 

[ZedClampet]

Uh, yeah Audacity has never had that capability.

It doesn't support that capability, but it does it anyway. What you have to do is select a small portion of audio and set it to loop and let it play for awhile. It eventually changes some or all of what is said. I have a video of it doing it to a recording I made. I was going to try to sync up a small portion of a video and I just let the recording play over and over while I worked on the video side. It was only about 10 seconds long, and it ended with the woman laughing in a sort of gravely way, and suddenly the recording changed the laugh into the woman saying, "I'm the devil", which sounds like the sort of thing someone would code in as a prank. In another part of the recording, it actually changed a curse word the same woman said.

This happened last summer. It's possible the code as been removed by now. I noticed the last time I was in Audacity that the entire looping process has been changed.

 

[Frag Maniac]

It doesn't support that capability, but it does it anyway. What you have to do is select a small portion of audio and set it to loop and let it play for awhile. It eventually changes some or all of what is said. I have a video of it doing it to a recording I made. I was going to try to sync up a small portion of a video and I just let the recording play over and over while I worked on the video side. It was only about 10 seconds long, and it ended with the woman laughing in a sort of gravely way, and suddenly the recording changed the laugh into the woman saying, "I'm the devil", which sounds like the sort of thing someone would code in as a prank. In another part of the recording, it actually changed a curse word the same woman said.

This happened last summer. It's possible the code as been removed by now. I noticed the last time I was in Audacity that the entire looping process has been changed.

Yeah that would definitely be suspect code in Audacity. Hopefully someone fixed it by now.

 

[Pifanjr]

uncredited, unpaid use of someone else's voice

Can a professional voice actor's voice be copyrighted? Their voice is their livelihood, so they should have a say if, or how, it's going to be used, as well as receive adequate monetary compensation.

Personally, I hope we can get to the point quickly where we admit that too much of human labour can be made obsolete with technology and we should do away with needing livelihoods, instead of trying to regulate technology just so we can artificially keep the need for jobs alive.

It doesn't support that capability, but it does it anyway. What you have to do is select a small portion of audio and set it to loop and let it play for awhile. It eventually changes some or all of what is said. I have a video of it doing it to a recording I made. I was going to try to sync up a small portion of a video and I just let the recording play over and over while I worked on the video side. It was only about 10 seconds long, and it ended with the woman laughing in a sort of gravely way, and suddenly the recording changed the laugh into the woman saying, "I'm the devil", which sounds like the sort of thing someone would code in as a prank. In another part of the recording, it actually changed a curse word the same woman said.

This happened last summer. It's possible the code as been removed by now. I noticed the last time I was in Audacity that the entire looping process has been changed.

I'm curious whether this might be a case of the human brain just doing weird things when listening to an audio loop. Human brains have a tendency to do weird things when processing sensory input in certain circumstances and it wouldn't be surprising to me if listening to a 10 second loop for a significant amount of time is one of those circumstances.

 

[ZedClampet]

I'm curious whether this might be a case of the human brain just doing weird things when listening to an audio loop. Human brains have a tendency to do weird things when processing sensory input in certain circumstances and it wouldn't be surprising to me if listening to a 10 second loop for a significant amount of time is one of those circumstances.

Sadly, after I posted that I realized that my video was on the laptop that fried. I've still got the original audio recordings on my phone, though. I might try to recreate the video. The laughter turning to "I'm the Devil" seems plausible under your explanation (even though I'm not sure that's what happened). If I just experienced that one incident, I definitely would agree with you. The recording was poor, and I didn't do anything to try to replicate or verify what happened.

The audio on the curse word, on the other hand, was very clear, and I turned the player off for awhile (somehow hoping it would change back), and when I turned it on again a few minutes later, she was still saying the word '$h!t' instead of "the f word", which is what she actually said, and I'm not sure my brain would mix that up, especially after having it turned off. Also, knowing what she had actually said, you would think that my brain would be predisposed to hear the correct word.

 

[Brian Boru]

I've spotted weird things NPCs do in games, I get a little creative dubbing in game dialog to go with

Hilarious! Great job

 

[Frag Maniac]

Hilarious! Great job

Thanks Brian, it's a short clip, but took a fair bit of time to make.

 

[Pifanjr]

The audio on the curse word, on the other hand, was very clear, and I turned the player off for awhile (somehow hoping it would change back), and when I turned it on again a few minutes later, she was still saying the word '$h!t' instead of "the f word", which is what she actually said, and I'm not sure my brain would mix that up, especially after having it turned off. Also, knowing what she had actually said, you would think that my brain would be predisposed to hear the correct word.

It might have been a Laurel/Yanny, gold-white/black-blue dress type of thing, where even if you know what it should be, you can't always force yourself to hear/see that.

 

[ZedClampet]

It might have been a Laurel/Yanny, gold-white/black-blue dress type of thing, where even if you know what it should be, you can't always force yourself to hear/see that.

Not really. Not for the curse words. The laughing, yes. But that's not what happened in the other one. After thinking on that more last night, I remembered exactly what happened, which we should be able to find in the post I made about it. What she said didn't change until I changed the pitch. On that one, it wasn't a case of hearing it over and over until it changed. I played it and she said one thing. Then I changed the pitch and she said something else. There was no ambiguity, and the words are not similar.