Anyone who dabbles in 3D software like Blender or Houdini knows that, despite the great strides of engines like UE5, real time rendering (and rendering in general) still has tons of room to grow, both in software and hardware. Proper real time simulations of light, fluids, etc. are only just now starting to become accessible, let alone ubiquitous. There's still plenty of frontiers yet to reach.

I think it will start taking 12 years or 2 gens to see notable impressive graphical jumps, we are kinda of really starting to slow down.

I think they'll be saying this about the Playstation 12 and the Xbox Series Fun, and they'll still be wrong.

All I know is I can't wait to watch a Photorealistic 3.5 foot tall Mario jump 12 feet in the air and see his little mushroom bulge flop around in his overalls. In 120k.

I have been hearing people ask this since Mortal Kombat in arcades since 1992. The answer is always. Yes. They will and the limit is when it's fully indistinguishable from reality. Not just in looks but in interactivity. Physics. Sound. Smell. Taste.

I would have said no cause of the cost it takes in money and time had I not seen Unreal 5 and the potential there. Ai also will help take the brute of the work that is now costing so much and taking so much time. Yeah, they can evolve but the question is should they approach realism or not, I think not. I think the direction to go is more stuff on screen, more particle effects, more dynamic stuff like Ghost of Tsushima. I also think if GTA6's trailer is a representation of it's graphics then that's where the graphical ceiling should be, look more than good enough.

Why would they need to get better than this...?

It's the perfect cap, just keep adding particle effects and atmospheric stuff to this and it's roses. You can see this in the part with the alligator in the store, it has atmospheric effects which greatly increase the visuals. Make these atmospheric effects dynamic and it's golden sauce man. Improve the lighting with Ray tracing and so on. I prefer the FF7 remake character models faces too where they're slightly anime in the eyes cause they don't give you the uncanny feeling. Animations, particle effects, geometry, clipping, detail pop in, shadows, contrast are all areas that need to be evolved but graphics on the surface, nah, jaggers are a thing of the past and we're already past the point where it's getting too real to be enjoyable visually in some cases like that body cam footage FPS game. I don't wanna look at real life on my screen, I want vibrant particle effect filled worlds like Ghost of Thsusima. 

Can they improve? Yes.
Do they need to improve much beyond today's high-end graphics? No.
Will graphics improve by leaps and bounds from here on out? No.

While yes people have been saying graphics can't get much better since two gens ago, we are in fact now at least getting close to that point. Sure over the next 50 years they will get better, but it's gonna be slow improvement and it won't really be worth it. Two gens ago to last gen was still a decent jump in graphics, though not huge or anything. But difference from last gen to current gen finally got pretty small. It's only going to get smaller in the future.


Graphics are already hyper realistic, and graphical improvements don't actually matter much at this point. I see improvement on effects (like full on ray tracing) and running all those nice effects, and improving draw distance and just in general the number of things happening on screen, while running all this at 60+ frames. But I don't see the actual detail of graphics improving a ton from here on out, and there is certainly not a need for it to improve. Techniques might improve like better facial movements and whatnot, that make things look more realistic, but actual graphical detail in terms of polygons in an object and textures don't really need to improve at all and won't improve much from here on out.

The industry is already at the point where making today's hyper realistic graphics causes games to cost an enormous amount of money and time to make. This needs to not get worse. Graphical detail doesn't need to get better. Higher power in the future should just be used to run the full effects with more draw distance with more things on screen at smooth framerate and keep improving AI in games, but not much more graphical detail.

I expect several system gens from now we'll still say PS5/XboxSeries look good and the detail of the graphics is still comparable, but that there's just fewer special effects and stuff happening on screen. At this point I basically consider all games to have good graphics, assuming the games are done well. Whether we're talking Switch or PS5, they all look good. Improvements in graphics don't affect how good a game is anymore.

So yes I agree with the OP's friend that graphics won't improve much, and furthermore they SHOULDN'T improve much or else every AAA game is gonna be taking 10 years to make and any AAA game that doesn't sell at least 10 million copies is gonna be a failure because they cost so much to make. Extra processing power in the future will just go to making basically today's graphics with more things/effects happening while still running smoothly, that's it. The difference between PS4 and PS5 already is fairly small, and each gen from here on out we're basically gonna say yeah I guess it looks a tiny bit better. In terms of graphical detail I think we're >95% of the way there now.

If you look at that GTA6 trailer, yeah in 30 years those graphics are still gonna be good and look comparable to whatever games are coming out at that point. Sure digital foundry might point out all the tiny differences between that game and GTA8 coming out in 30 years (only at 8 cuz these hyper realistic games take so long to make), but to the average player it's gonna look pretty dang similar.

Everyone and their mother in the industry knows there are at best a handful of full nodes left and then they are moving on to vertical circuit stacking. This isn't exactly a secret or an opinion:

Regardless of TSMC's shenanigans, ITRS's nomenclature has never referred to the physical size of transistors. Historically it was half of the gate pitch which was a close approximation for the smallest feature size of a MOSFET.

That hasn't been the case for a long time, however, as logic scales better than the rest of the circuit:

TSMC's 3 nm transistors have logic gates of ~ 15 nm and within them transistors with silicon fins as small as 7 nm wide, which is pretty close to electron mobility collapse from quantum tunneling.

Even using theoretical 2D materials and 1 nm transistors as wide as a single molecule isn't going to make matters much better (just ~ 5 times denser than TSMC's N5).

Everyone is trying to find smarter solutions for computing problems or trying to find quantum algorithms that actually work (hint: they won't since it's very very very likely that P =/= NP) because they know brute forcing with MOSFETs isn't going to cut the mustard for much longer.

And I'm just imagining the absurd power cost in watts this will all need in the future, just because of our obsession of achieving the true simulation of our real daily life.

Actually, looking a the inner workings of how UE5 renders, I think we have or are most likely to achieve surrealism in the process.

Anyway, in the truest sense to me graphics rendering to the "realistic" style have gotten staller and more complex the closer we get to it, while also not bringing the added complexity needed for it to be spectacular.
Anyway, diminishing returns have seriously set in since the last generation and the biggest noticeable difference came in from rendering resolution support like DLSS, FSR and frame generations and the shortening of loading speeds to almost nothing. Arguably though, that's just something we already had back in the 3th Gen/4th Gen when games were mostly cartridge based.

And then you consider the money and time cost(dev time) of it all and realize it's grown exponentially while the audience hasn't. That arms race chase for "realism" is costing us far more than it should at this point.
I think what we achieved so far is arguably satisfactory and there's other aspects of the video games experience that should be prioritized instead of graphic rendering.

Developing new artstyle, advanced AI, more impressive physic simulations à la BOTW/TOTK, etc ...

To add to this part something noteworthy is that Rift Apart actually looks better than the 2016 Rachet & Clank film in certain ways despite only coming out 5 years later. This is due to the film having a modest budget but it shows that the days where a game like that can have similar visual quality to a recent Pixar film are getting closer.