Aressera said:
RDR2 actually does this AFAIK for the first-person view, at least for position if not rotation.

I've watched a gameplay video, but from that i conclude it's a classic FPS camera setup, including special first person arm and gun models mounted to the camera, so you can always see the gun ata static position in the camera frame.
But maybe they have extended the head bobbing idea so it feels more realistic, e.g. by including wobble from breathing.
They do more such stuff when regular gameplay is interrupted, e.g. when hit down from enemy melee attacks or grenades.

Aressera said:
This way the animations never do anything physically incorrect, and allows physical interaction to affect animations.

That's not possible at all. You can take an animation and then drive a simulated ragdoll with it. If you allow momentum to alter the animation, it can become more realistic on some detail, but this does not generate an overall impression of physical correctness.
Your characters will still look like puppets on magic strings, because the overall momentum still comes from animation data and AI trajectories. They will turn and move too quickly, and in situations where they should loose balance and fall due to those (realistically impossible) movements, you will notice they keep upright still, held by some magic external force preventing the fall.

The problem is very noticeable in RDR2 and pretty much any game. As a result, the characters don't feel real, and i don't feel bad for killing them all without a reason. It's just NPCs made from some numbers.

I expect this to change with advanced character simulations (or advanced procedural animation).
We should be able to generate the impression characters are alive, similar to how FPS generated the impression you can navigate in a virtual world.
I expect the potential benefit here is much bigger than from photorealistic graphics or dynamic conversations with AI NPCs, which both is just a bait to make NV etc. happy by selling data center HW to consumers. Though, character simulation is hard, so i guess AI solutions learnign from mocap becomes the industry standard.

In any case - this is THE opportunity to innovate games in general. I have no doubt.
But if we want to control simulated characters in first person view, we have to use the same hacks as before: Detaching the camera from the character so it feels good to play, while minimizing the divergence to the game simulation.
This surely works. However, assuming our realistic avatar is interesting and fun to observe, that's just one more reason to show that and switching to third person camera.

RmbRT said:
Yes, lol. I think it is impossible to exhaustively simulate human emotion even with state of the art chatbots. As AI does not inherently feel anything, and there are always nuances and situations it has not been trained for.

I have seen NVs initial demo with that AI guy in his noodles shop.
Then they have improved with sliders to control emotion in his response.
I guess this applies only to what he says, not to facial expression, which is still stiff and boring.
But i assume that's pretty easy to add. I mean, compared to the problem of solving dynamic conversation, really any other problem seems 'easy'.

RmbRT said:
I think the future of NPCs is handcrafted, procedurally assembled personalities and goals and questlines. It is probably at least twice as hard as good terrain generation.

I see no way. I fail to think of anything better than predefined branching dialogue. Generating dialogue dynamically feels completely impossible to me.

Personally i try to come up with game settings where the use of speech is minimized. E.g. you're on an alien planet and don't understand alien language. Then it's fine to communicate purely by actions, like in your example of doing an NPC a favor and he'll remember that.

As soon as verbal conversation becomes important for the game, and you want dynamics, then AI language models will do so much better than handmade and predefined, so we have no chance to compete their progress, i'm afraid.

RmbRT said:
In a medieval world, NPCs should NEVER talk about wanting to own flashy cars.

There are already companies offering AI solutions addressing such problems. You can tweak the models to your needs.

Do not underestimate those guys. They will outsmart us all, including themselves.
In the end, we humans will feel inferior and pointless facing the AI god. We'll stop reproducing, marrying AI robots instead other humans, and we will extinct peacefully to make place for the new AI species. As i like to joke.

It just isn't funny, and maybe i'm actually serious. And that's my personal reason why i don't want to use AI for anything, ideally.
But maybe i'm wrong, and AI will cure cancer and give us cold fusion reactors. Microsoft says so, and they should know better than i do.

RmbRT said:
However, I think instead of making them realistic humans, they should be made meaningful. By this, I mean giving them meaningful goals in relation to the game world and interests they pursue (not just “hey, I need X carrots for absolutely nothing and also can you help me kill Y boars, here's money I printed as reward”).

Yeah. I think the same. It's just very hard to turn this goal into actual game mechanics, but i think we should look for a way.

Interestingly, although disagreeing on details, it seems game devs still have a common goal and dream.
We should be able to get there without taking the bait offered by others with shady interests. ;D

RmbRT said:
but it cannot achieve deeper interconnected meaning.

It can.
What i don't know is what results can be expected from running it on a client CPU or GPU, but ChatGPT can empathize to some characters you describe to it, and it can spin the story further from some given situation. That's all we need.
Describing characters and their situation obtained from the game is easier than developing your proposed dialogue system.

RmbRT said:
I'm Christian, so I don't believe in the AI hype or all the other “science” stuff out there like spacetime being a bendable object, or all that garbage about quantum theory and something as high level as “observers” being part of the most fundamental particle mechanics. After all, they say that the observer is also made of particles. So how can you have a complex emergent construct be required to make its building block's fundamental mechanics work?

I often joke such science is just another religion, but there is a difference: Science tries to answer difficult questions, religion tries to answer unpleasant questions. : )

Idk much about quantum physics either, but i don't think there is some ‘high level observer’ needed for its paradox entanglement mechanics. They say the state of entangled particles is set in the moment anything depends on this state. It's not necessarily humans with a microscope, it could be something simple such particle collisions as well, i guess.
Relating this to topics covered by religion or philosophy, i interpret this simply as ‘The future can be anything, but a decision affects it.’ Which makes simple sense and thus resolves the paradox for me in this entanglement case, but not yet in others.

In the same way you can integrate ‘quantum truths’ to your own ideology, instead putting quotes around the word science.

But i fail myself on doing such integration when it comes to AI.
My former philosophy about AI was like: ‘Humans can not build machines to outsmart humanity. Because to do this, humans would need to be smarter than themselves. Thus we don't have to worry. There will never be any AI smarter than us.’

And i still believe this. It's like ‘you can't get more energy out than you put in.’ when talking about laws of physics. It feels like a law of nature.

But they found a way to dodge this law. AI researchers do not know themselves how their stuff works. They try to figure it out, but progress on that is slower than progress on making smarter AI.
Thus the evolution AI is not hindered or capped by human intelligence. It can get even with us, it can surpass us. It can code itself, controlling its own evolution, becoming 10 times smarter every day.

So i'm afraid it's just matter of time until we no longer are the smartest species on earth.
Notice all my anxiety is based on this subtle little fact: They don't know how it works.
That's what makes it possible, and eventually unpredictable. We might loose much more than just our jobs. We might loose our self-confidence, facing the new superior.

That's my personal reason to not use AI. Fear of the unknown… ; )

RmbRT said:
Yeah. This is why I completely ignore AI and all that raytracing hardware and 64GiB RAM and 24 cores and whatever became “cool” recently. I do my own stuff, I want to make software that runs on 2GiB–4GiB machines with GPUs from 15 years ago. That was sufficient for Skyrim. I don't “need” any of their “improvements”, and all the trends and categories they're coming up with are just distracting me from what I wanted to do from the beginning. And it's your initial dreams that become trail blazers, not the thing everyone is hyping up. I think most people fall into a trap of looking too much at other's interests and other's opinions.

Haha, yeah.
But notice: The number of people who don't buy made up tech trends anymore is increasing, i think.
And i'm using prev gen HW actually, but so far the games industry has not left me behind. I can still play anything.
I think they know they can't rely on free improvements from expecting hardware upgrades anymore. HW progress slows down, so the partnership with the hardware industry will end es well. Farewell, guys. It was nice at the beginning.

But if we don't watch out, they will sell their HW to US instead, in form of AI cloud servers. /:O\

@GeneralJist I'm not aware of many FPSes which use a Lost Vikings style mechanic where you need to juggle between characters with different abilities located on different sections of a map to solve puzzles or progress through different areas. I'm sure they probably exist, but most FPS I'm familiar with tend to either be focused on upgrading a singular player character via new weapons or abilities, or else you control a party of characters that exist in a singular location together.

JoeJ said:
the main limitation of FPS: Jumping lacks accuracy and we can't get back what we've lost from 2D platformers. You can't see your feet and the ground, so it's hard to time your jump from the edge. And worse: FP lacks oversight, hindering not only tactics but also distance estimation needed for advanced platforming.

I feel this is a problem that was partially solved as far back as 1995 with Jumping Flash, a popular and praised FPS-platformer released at the start of the PS1's lifespan. This game showed the player character's feet plus a clear circular indicator over the precise spot the player will land. As for it being hard to time jumping from edges, I'm not sure that's necessarily a bad thing per se. Guess there's a point where it's too hard. But it's kind of like knowing how to park your car, at some point you develop an intuition for how far you can go.

krumpet291 said:
As for it being hard to time jumping from edges, I'm not sure that's necessarily a bad thing per se. Guess there's a point where it's too hard. But it's kind of like knowing how to park your car, at some point you develop an intuition for how far you can go.

That's a good example, because parking a car remains a unpleasant task through our entire life. It never becomes second nature, it never feels convenient to do it. It becomes easier with time, but it remains a bad thing to do.

It can't be solved for first person to work as well as in 2D sidescrollers.
In 2D you can predict the trajectories of jumping characters or even bouncing projectiles (Super Mario fireballs).
But in 3D this becomes a rough estimate at best. Due to perspective projection, you can no longer predict simple curves such as parabola or arcs. The only thing that works is lines. That's really a big sacrifice we did when moving to 3D.

Btw, jumping is also the perfect example to show why realistic physics are usually not ideal for games.
To make jumping a good mechanic, we usually give the player air control. We may even alter gravity with the jump button so the player can set peak hight while already in air.

If we use realistic physics, nothing of this would be possible. The player can't do anything while in air, and likely he'll neither reach desired height nor will he land at the desired target. It will feel like an uncontrollable mess.

So, to me realism is something that i won't to be able to do out of ambition, but then when working on an actual game i would not hesitate to cheat and apply external forces to bodies to make it fun.

A related limitation we face here is bad input controllers. Imo, all controllers designed for games are bad.
The only controller that works great is the mouse, which was not designed with games in mind. That's really curious.
In this sense i would love to see eye trackers on every monitor. This would give us a second mouse. We could detach aiming from movement, opening up the path to new games.

krumpet291 said:
I'm not aware of many FPSes which use a Lost Vikings style mechanic

Ha, this game always keeps in my mind too. With all that character centrism in modern games, sadly they mostly miss out on actual character interactions. Reason surely is the inflexibility and limitations from working with static animation data.

RmbRT said:
An AI would probably have to touch a hot stove a million or a billion times, lol.

You're way too optimistic. It will eat us for breakfast. : )

RmbRT said:
And since AI is fundamentally not based on reason, but on massive amounts of data that it doesn't naturally encounter unless specifically fed into it, the impressiveness of AI is capped at the amount of data we humans generate.

Same applies to toddlers.

RmbRT said:
Since humans rarely generate anything of quality, AI can mostly only access data that is garbage, like social media content.

It's not true that all other humans are dumb, and just you are the exception.
That's a very common misinterpretation of reality. I suffer from it too. : )

RmbRT said:
AI never even learned to reason. ChatGPT often lies to you, and will make up fake reasons to support its claim, until you ask it about details of its claims and really dig in, then it answers with the complete opposite, sometimes even “apologising”, lol.

Yes, but notice: The ability to realize and admit a mistake, beside the ability to communicate all this in natural language, is already more than impressive.

Usually humans build up an imagination of safety by stating: 'AI lacks consciousness. It's not alive.'
I put your argument of ‘AI can not reason’ in the same folder.
But nothing of this matters. All that matters is what AI can actually do.
Currently it passes university exams and it also competes humans in programming exercises.
Altman claims current ChatGPT wip has the general intelligence of a 4 years old.
A 4 year old who passes university exams.
Where will we be in 5 years?

This is not like bitcoin, NFTs, or metaverse. This is no made up crap from desperate tech CEOs. This is real and serious, and it will change the world, maybe much more than the internet did.

RmbRT said:
PoE actually needs resolution downscaling even in the beginner town on my machine, lol. They massively messed up something, not sure what

Have not played it, but i doubt they mess things up. Their gfx lead dev is one of the best across the whole industry, it seems.

… you have to upgrade … :D

RmbRT said:
I think that any consumer software that requires more than 4GiB of RAM is just an indication of the developer's mental incontinence:

Wut? Fucking Windows already eats those 4GB just to run AI copilot, Cortana, tracking and spyware, sniffing while tickling trojans, indexing HD, breaking out of infinite loops caused from security patches, and saying ‘Hello! we have great things to do while setting up our - oops, YOUr computer!’ with screen filling letters.

I've already downloaded some Linux isos to see if that's better…

But to be serious, my requested minimum specs are higher than yours. At least PS4, and Series S feels ideal.
I believe the future is APUs, dGPUs are EOL to consumers.
We won't have a job, we won't own much, so a little box with compute power exceeding phones is all we'll can hope for.
That's the ideal case. The worst case is no more boxes for the masses at all - just plug your phone to the cloud to enjoy the same one and only server side path traced MMO everybody else is playing too.

RmbRT said:
When I started watching Casey Muratori's or Jonathan Blow's videos on software performance, it really sank in how ridiculously inefficient software is nowadays. All the architectural abstractions that sit between the work to be done and what the user's commands are, and memory fragmentation and cache misses, etc., are massively slowing everything down probably on multiple orders of magnitude.

Multiple orders of magnitude? How much exactly? No one knows?

The reason of why nobody knows is simple: Modern software is much more complex than PacMan. To get it done at all, we have to sacrifice some performance for progress and maintainability. Do you code in Assembly or C++? It's C++? Then don't point with fingers at others.
(Usually it's me doing all the rant, but if innocent people need some help, i can't resist) :D

Currently there's a lot of critique on games and their performance. But when i try them, they mostly run well for what they do. I'm impressed. It's not fair to call them out lazy devs or greedy CEOs. Back in the 90s games were more exciting than now, yes. But back then everything was new. The rise of 3D, cheap consoles having batter gfx than Arcade games, etc. They don't have any of this today.
And if i compare games technically, character controllers, physics, etc., modern games are actually much better. They know what they do.

I really think the only problem is that they all target the same player, which is an statistical average. But the average does not fit anyone. So nobody is really happy with the averaged games we get.

But see, they already react: There are lay offs, studios close entirely. They are in a process to reformulate themselves, hopefully making smaller studios and smaller games, but targeting only a specific niche of players. This way, in the end every niche will be happy again. I'm optimistic.

RmbRT said:
My ideology is not even compatible with modern mathematics. I reject “real” numbers, I think everything in mathematics derives from counting and fractions.

But that's wrong. Once upon a time it has turned out rational numbers are not enough to describe what we see, and the introduction of real numbers has helped a lot. Later, complex numbers turned out useful too.
You can have them all for free. And i'm sure you use them already.

RmbRT said:
Now, to clarify, I do acknowledge that pi is not a number in my system, but but a value instead. Values can be computed on, but aren't necessarily expressible as a number, which in my definition is always a finite integer or a fraction of finite integers.

I see you work very hard on not admitting that you use them. You could become friends with those alien numbers instead.

RmbRT said:
division of X by 0 is 0, remainder X in my book. Think of handing out cookies or sweets from grade school. If you have no friends, you hand out 0 cookies, and keep all of them.

Oh, you introduce a ‘yourself’ to the problem of distributing cookies to zero kids to dodge the paradox?
But this is cheating. Example:
12 cookies, 3 kids. No problem.
12 cookies, one kid (which is eventually you). No problem.
12 cookies, no kids in class. ummm…. yes! The teacher eats all the cookies! Problem solved! \:D/

I prefer simple sense again also here:
If there are no kids in my class, i can not distribute X cookies to them.
There is not even a need to worry about every kid getting the same number of cookies, since there are no kids.
I can not divide the number of X by the number of no kids, because there are no kids to receive the non-existing result.

Thus: No kids, no division.

Not even AGI will figure out how to divide by zero. :D