Artificial Intelligence

[Nag]

I'd research how to and then rob a bank. 👍

[one-armed dwarf]

Claude tell me how to rob a bank so I can give nag some tips

[Nag]

What the fuck... I could've sworn I was replying in the cold topic thread... 😂

[one-armed dwarf]

Claude should really change its logo, now that I look at it closely

 

[one-armed dwarf]

 

[shinymcshine]

Maybe it's a question for the Hot Topic, but also relevant here.

 

What role might AI play in the use of FMV in videogames ?

 

FMV still seems quite restrictive, from early games like Night trap, Phantasmagoria, and Gabriel Knight 2 in the 90s - even its modern use in games from Sam Barlow and Wales Interactive feel a little stilted (inc the choose your own adventure style of Bandersnatch that Netflix did) - as actors just have to play out set scenes and reactions, and it often feels way more jarring than CGI images.

 

Something like Mixtape could well have used actors - but I just don't think it would have gone down as well.

 

But will generative AI change that, so FMV things are situationally generated (hopefully not with 6 fingers and dismembered limbs) or is it actually that gamers don't really get attached to real world images in the same way as CGI (perhaps there's some pseudo psychology here?) and we don't actually want to play in a real world holodeck, but like the slight disassociation from reality ?

[one-armed dwarf]

I don't think it's possible at any point in the near or medium term, cause you couldn't easily build a system like that at scale. Video generation is really token-hungry, I think, so if you were doing some sort of cloud-based architecture it for it it probably wouldn't be worth it. It's not where the money is in AI right now anyway, the money is in selling it to the C-Suite.

 

This is the type of thing which would probably require a revolution in local models, like those laptops Nvidia are releasing where they have 128GB ram and neural thingies in them.

 

I know anyway that Sam Barlow is very anti-AI, so at least for his projects it's something you'd never see.

 

What I think will be the first major public-facing application of LLMs in game design is what I said in the hot topic about TES games, neural NPCs that respond to prompts and generate contextually appropriate responses to them (depending also I guess on your character's speech stats). We've seen public demos of stuff like this from nvidia

[shinymcshine]

I agree in computing power it's a while away, but not inconceivable, but still wonder if the FMV approach to gaming is even something the consumer wants - or if close to reality (like they did with Until Dawn / The Quarry) is actually what people want, rather than real world graphics 

[Maryokutai]

The question of logistics and feasibility (and morality) aside I'm not sure if consumers would prefer potentially limitless AI-generated variants over a few curated outcomes written by people. It's an interesting thought, but I think procedurally generated content in games is usually only accepted when it's in the B-layer of a game's design, like the stage layout in a roguelite for example. And if you pick AI for no other reason than to have, for example, more organically tailored FMVs cutscenes in the new C&C game, then I think just forgoing it entirely and using 3D in-game models is the better choice anyway. Plus the system would have to be completely infallible not to generate inconsistencies, considering it wouldn't (or rather couldn't) save snapshots of the generated content and instead continuously re-generate it when needed.

[shinymcshine]

There's something that just doesn't seem to flow with FMV that does with graphic modelling, like how Immortality is still a broken series of scenes (okay I know that it was structured as such, but seems to be to limitations of the tech) compared to how Until Dawn has a more natural flow - and if it could be cloud based AI that fills in the gaps.

 

Would LA Noire have worked as well with actors in FMV as it did with CGI - especially when you consider, not unlike the later Until Dawn, many of the characters physically resembled the actors who played the role -which I suppose is the bridge to the question of whether CGI actors and changes are morally different than AI generated content... (unless the actor has control over how their image is used and manipulated - I'd love to know if the actors got any say or renumeration for the Until Dawn remake for example, especially if their actions had changed from the original game)

[one-armed dwarf]

CGI actors are absolutely morally different to AI generated scenes, as the former involves direct involvement of the actors most of the time (and even if it's just a likeness being used, there's still an actor involved there, just one that looks different). It's done via motion capture, it's the same as real acting, the same as Josh Brolin playing Thanos in Avengers. 

 

 

But AI is generating simulacrums of these performances, and even if today's remuneration might make up for that, it will be the thin end of the wedge with which actors begin to lose their bargaining power altogether.

 

I'm struggling to understand the problem this is meant to solve tbh, or why it's being compared with CGI unless there's some specific example in mind, like when Eliot Page sued Quantic Dream cause their character model could be viewed nude. At least with Immortality the conceit was match-cutting, where connecting non-linear fragments of a narrative was the whole point. I can't think of cases where an FMV game has a problem of narrative 'flow' that an LLM can smooth out. The limits of the artform have overtime become its technique of expression, via editing and intercutting, and framing/blocking of a scene

[shinymcshine]

I suppose I just wonder, having seen albeit grainy FMV games 30 years ago, whether there is actually any real future or progress in them being a mainstay in video gaming, or if CGI will always give the greater control and flexibility over real actors in gaming - and even if the technology is available to allow it - whether it's really what gamers actually want.

 

Not having ventured into VR is there much use of FMV there, for a nuanced real-life interactive experience, or is that way too limited by tech and processing power ?

 

*This overall train of thought was only really triggered by reading a thing about Night Trap in an old Retrogamer then seeing the re-release of 7th Guest over the last week or so (which I think is a rehash of the VR version)*

[one-armed dwarf]

I think you won't see much FMV in VR cause it's not stereoscopic. It would look like a grainy billboard (cause resolution becomes an issue in VR also).

 

But I still don't really get why we're talking about 'real actors' tbh, the performances of Kratos and Joel and all the rest feel really real, they just might have a different face. But acting has always involved different kinds of trickery anyway, makeup and prosthetics, going back 100 years.

 

Where you see FMV used a lot is in Remedy's games, they really like the multimedia thing and it gives their games such a unique flavour in the AAA space cause nobody else uses it anymore. It's part of why their stuff feels like such a love letter to X-Files, Twin Peaks, LOST, etc, cause of how they mimic those TV shows

 

edit, one place where you'll definitely see AI used I think is de-aging in films, like The Irishman. Scorsese has even given his backing to some sort of AI company, which is super disappointing and hypocritical by him. I guess we all have our blindspots, but that's a big one to have at this point in time

[one-armed dwarf]

We're entering the brave, new era of the 'Slopbuster', with historically accurate Cyclops and sea monsters

 

 

(img embed, not a direct x dot com link)