All of them synthetic, lacking author's touch and unique personality. No thanks, I'd prefer a few masterfully created characters by a talented game writer.

Removing the need for humanity is a novelty ATM. How cool, right? But we'll soon see the world that eliminates the need for human expression (replaced by plastic mimicry)... and then the need for humans almost entirely.

And what happens when most people are no longer needed...?

Genocide, then probably perpetual wars by the machine masters. Like Reign of Steel, maybe.

AI writing, AI voice acting, AI-generated faces and AI NPC behavior. Sounds like Oblivion NPCs v2.0, I can't wait for the memes.

So basically an advanced speech synth, making more lifelike voices from a text input than "normal voice synths"?

I tried it with a couple of texts, and yes I really couldn't tell it wasn't a real human. Depending how much CPU processing that takes, it would be a good way to severely reduce the size of many dialogue-heavy games which have lots of recorded voice acting, and naturally they'd be much cheaper to produce too, as no human voice actors are needed.

All in all, great stuff, if it really is "just" an advanced speech synthesizer. How far we have come from the Texas Instruments TI-99/4A times...

https://www.youtube.com/watch?v=kUPkCNcT1Yw

So... is the meaning of life... to work? Otherwise life is meaningless?

System Shock and the movie "Elysium" had those cool devices which would heal the patients of any ailments in a jiffy.

https://www.youtube.com/watch?v=f4gmgTebHog

If such devices were real, should they be destroyed so that doctors and nurses would not become unemployed?

Back in the day, lots of human call operators connected manually all voice calls to their destinations. Now they are all unemployed, and computers connect us. Fair?

I personally think that the less people need to work, the better. Sure there are problems when suddenly some group of people become unemployed, but they just need to find something else meaningful to do in their life, like golf, TikTok, drugs etc.

So, why were good old midi's discarded in favour of pre-recorded music? And for what purpose do even some indie studios hire real orchestras and bands to perform soundtracks for their games? ;)

Obviously, I am not referring to main/secondary characters. I was thinking of the silent, robotic, lifeless NPCs we still have in open worlds. Do you really think this technology isn't an improvement over that? Imagine the possibilities: conversations between NPCs, radio DJs, TV news, followers, merchants, etc....

Simply a bad design, not a lack of resources. This problem has been solved in Gothic 1.
I prefer pre-recorded broadcasts, more or less tied to the story ( like in VtM: Bloodlines )

***

AI-generated voice could take the same tasks in games, for which we use it in real life. Public address systems in shopping malls / on transport, "talking" vending machines / info stands, etc. And of course, the voice-over of automatons and AI-systems themselves. :)

One other use: In an RPG, use this for battle narration. For example, if an attack deals 1743 damage, you could have the AI voice actor (based, perhaps, on a sports broadcaster) shout "one thousand seven hundred and forty-three damage!", and similarly for other things, including character names.

(I'm thinking of battle narration like you see in games like Dragon Quest and Wasteland 1, here, but voice acted.)

I listened to one of Jerry Seinfeld, written by an AI, performed by an AI, and it could easily have been him reading a script as a dry run. The jokes lacked punch, but it really could have been a pre-show read through. I was an AI though. And the act wasn't terrible, just not great.

Considering GPU's are primarily hundreds (or thousands) or simple CPU's that don't have privileged access and primarily do math, not too unlike offloading work to the FPU back in the day. Though how big a chunk of code and what it can do i am not sure. And as long as you have enough for both, i don't see why not.

I'd honestly like to see more programs trying to take advantage of the GPU. Say lzo/zlib compression/decompression, image optimizations, more video filters in ffmpeg (like frame interpolation), encryption/decryption.

GPUs are good at one specific thing: Performing the same operation on a large amount of data. This is good for some tasks, like computing the Mandelbrot set or similar fractals, and for image processing, but won't work well for things like checking whether some large number satisfies the Collatz conjecture.

Well, they're also really good at rendering triangles, but that's not as useful for computation. (Anyone know about how many triangles a modern GPU can render per second?)

Edit: Another issue is that programs need to code specifically for the GPU, and may need to write entirely different code for different GPU manufacturers, as Intel, AMD, and Nvidia can't agree on a single, fully functional, standard. And then you get ARM devices, including smartphones and recent Macs, which have their own graphics hardware and, for example, don't support full modern OpenGL.

I tried some snippets from old game reviews that I've written. They had some capitalized words in there and it was impressive to see how the software structures the rest of the sentence around giving extra emphasis to the capitalized word. I see a range of applications for this great tool. "Voice acting" is of course none of those.