What will Unity be like in 5 years time I wonder...

It's not about the tool, it's about the company maintaining it.

 

My expectations for Unity in 5 years:
- Mono to .Net Core Transition will be done (thus Script Performance / GC should be much better)
- UI Toolkit will have fully replaced Unity UI and the UI Performance / Ease of editing UI will be improved
- URP / HDRP will have replaced the Built-In Pipeline
- DOTS will be finished and ready to use (however MonoBehaviour workflow will still be the main workflow for most games)

 

It is a matter of hardware, because living word means things happen even when the player cannot see them.
If your npc takes 1 millisecond to process, it'll take 3 hours to update 12 million of them.

On top of that, modern games are shy when depicting crowds.

This sort of thing is impossible:

Spoiler

That's humans.

Each having a mostly unique texture, unique goals, unique inventory content.

There are multiple billboarding tricks everybody is aware of, and instancing, but that's not the "living world " level.

Because in reality full freedom game means that in such crowd, you can approach any character, ask them anything and potentially convince them to join you on an adventure. Or any such character can approach you and start an "adventure" whether it is good or bad.

All of them are wrong tools for the job, in my opinion.

But Free Guy or Westworld, is pretty much what I"m describing. Koreans are also dreaming about VRMMOs with huge living worlds. That's also not a thing.

Actually, we have space engineers with destructible voxel-based baby planets.

Spoiler

Notice that this is from 2015.

You can build a super drill, start digging down and dig quite far, but below certain depth the game will simply delete your drill and anything you drop there. The limit is about 7 kilometers below the surface, I think.

 

Games are smoke and mirrors. For example your crowd example. You have a crowd system and when you approach a individual in that system that individual is removed and replaced with a much more capable NPC

 

I'm more optimistic about Unity now than I was last year. Lots of room for improvement, but I can slowly see the pieces coming together.

 

If a tree falls in a forest and no one is around to hear it, does it make a sound?
I'll guess most developers would answer: "No it doesn't make a sound, why it should?" while most players would say: "Sure, it does make a sound even if I'm not there"

Point is, with an increase of processing power increasing realism of not apparent features is worth it?
My answer is that's a design choise but in the last decades games went definitely towards realism: elaborating a bit we don't really want features the player can't see or perceive but how many times games make a damaged car, destroyed building or dead npc respawns/resets its state as the player goes away? And when the player returns would instantly understand it and perceive as unrealistic. That could be accepted as normal "game behaviour" and maybe we are used to it but would not fit every game type specially for games that demand realism.

 

See Dwarf Fortress.
--edit--

Jokes aside, there's idea of "Emergent Storytelling". When you have a world, it acts according to certain laws and rules, and as a result of those laws and rules, unscripted story appears.

In practice if somebody would be able to make a sandbox with realistic one million npcs, that's something a player can spend YEARS playing. It would be a supreme version of sims. Or the perfect god game.

Imagine, basically, that you're observing one unique character in the world, it gets in a fight with a random stranger, then you can look through stranger's life history what led to this event. And maybe rewind/rewrite it. This sort of thing would have an incredible appeal.

However, implementing it is extremely difficult and expensive in term of storage and computational resources.

 

To be honest, I see this feasible now with DOTS support.
People already did 1mln entities that are moving and rendering.

Obviously for NPCs need a bit more than that.
But that is mostly AI logic.
However, neither they need update complex logic at 60FPS, nor it has to be 1m of NPCs.

More realistically I see 1k, 10k, or maybe even 100k NPCs, which are tracked through the story.

Factorio is a good example, where one thing can effect whole chain in the line. Just like some random NPC event.
And we have thousands of things happening. So such mechanics it is not out of question.

So it is more a matter of time, when someone do create massive SIMS like game
And it will be in Unity, more than other engine I believe.

 

Making 1mil boids fish swarm is quite different from having 1mil full blown autonomous agents with decision making.

I don't see this as feasible and will be extremely surprised if anyone manages to pull this off in any engine.

Sure, you can make 'particle crowds". But the point of living world simulation is that entities are not particles.

 

Yes, I believe games have still much to tell in terms of engineering and creative endevours and may need better hardware to fulfill them.

However if you are looking for emergent behavior/gameplay I don't think millions of agents are really needed but thousands should be enough, as Conway's Game of Life with hundreds of cells shows emergence and self-organization with simple rules. On the contrary typical boids simulations, you can do on the GPU, exhibit self-organization but relatively simple emergence. That's not what you are looking for, right?

I think you could write a simulation (if you could harness the power of the GPU it would be better than using all CPU cores) with thousands of agents and your own rules, starting with a minimal set of rules and adding a reasonable one until you get something unexpected. Boom! Emergent behavior.

Maybe it would not be as simple as described here. Sorry, couldn't resist, nevertheless if Dwarf fortress achieved that with hardware from more 10 years ago, today it's possible to do more.

 

That'll take 20 years, and it is called Dwarf Fortress.

The difference between what I describe and conway's game of life is that game of life is abstract and unrelated to reality in any form. While a world would require a story. With boid-like approach at best you'd be ablt o simulate biomes and t heir propagation, but then it will be extremely unstable. Basically, wildlife. And not a civilization.

There's actually another game that had a stab at a large world with a lot of agents, and that's Crusader Kings. Total number of NPCs is unknown, but some estimate them to be in 10s of thousands, however they are not "decision making agents" in normal form.

Regarding Dwarf Fortress, they did not achieve it fully either. There's a huge difference between story during worldgen and story when player starts interacting with it (either in fortress or in adventure mode).

 

So exactly what are you stating here?
We don't have the required hardware to make an emergent storytelling/gameplay game? Or we can make it but it's not enough? Or we can make it but it takes 20 years?

 

We don't have the hardware and algorithms.
Some notable early attempts exist, but they're not complete.
Dwarf Fortress is one such early attempt. And building another one like it will take similar amount of time.

For example, regarding your earlier proposed approach:

In practice it leads to situations that were seen many times during dwarf fortress development.

For example:
Athletics make one a dangerous opponent.
Swimming increases athletics.

Sounds logical so far, right?
The outcome:
River carps murder dwarves by dragging them off the shore and those carps are very hard to kill.

Why? Because fish swims all the time, that increases athletics. So you get murder carps of doom with maxed athletics.

Is this logical? Absolutely. Is it correct? Well, not really.

 

So, talking about "what will be unity in 5 years" i think if unity wants to be again in the eyes of everybody and retake the faith of their users, maybe an "industry changer" feature will be a good idea, i was thinking, imagine if you could run that "the matrix" demo from unreal, the one with the city but on a low/mid-end laptop at 60 fps, that will be definitely a game changer feature, something impossible, like lument or nanite 5 or 6 years ago, i know is sounds impossible but is not that what tech companies do? idk, just my humble opinion.

Now about the MMO think with millions of NPC:

maybe with LOTS of years of very hard work, maybe 20+ years yeah.

but also....

i agree with this

MFG 20 years?? i didn't knew about that game, thanks for the info

 

The devs worked on the game since 2002. It was originally an ASCII title, but there has been a r ecent graphical steam release. So, yes, 20 years in development.

The simulation aspect is incredibly thorough. Down to anatomy (creatures have layers, and not hitpoints, can die from pain, blood loss, or heart being destroyed)

Spoiler

The unfortunate thing is that this game also looks amazing in isometry (there's a fan made isometric visualizer), but the developer decided not to pursue this angle.

Spoiler

 

I mean, in 5 years you will probably be able to run the matrix demo on a low to mid-end laptop in Unreal, since current high end laptops will me mid / low by then.

So mis-interpreting your post a bit, what you're saying is, you wish in 5 years Unity will catch up to the Unreal of today

Personally, I think after they finish UIToolkit and its editor integration, they will get their first industry changing feature:

Unity will be the first software that needs 10 hours for the UI to react to the user's input.

Entering play mode will be counted in ut -> "Unity-time" and 1 Unity-time will be equal to 3.1415 hours. During Unite 2027, Unity's management will boast about how they finally delivered on their 500ut iteration time promise.

The will finally deliver their next generation animation tools, they will be a little slower and a little more cumbersome to use than the animator. Upon dissatisfaction by the community, they will instantly abandon the animation tools and start working on new animation tools that will take another 5 years to get into "Unity production ready" (Unity production ready is a new term, which means the feature passes the test of not crashing the editor when it's used for up to 5 minutes)

The progressive lightmapper will still not have feature parity with the beast lightmapper we had in Unity 4.

Each year, they will decide if they will still include Enlighten, depending on whether a groundhog sees its shadow.

The new audio stuff will still not be ready, the numerous issues and bugs to the current audio stuff will still be there.

Timeline will still be plagued with numerous "off by 1" issues.

Cinemachine will still be boasted as a great feature, even though every game that relied heavily upon it was panned by reviews for their terrible camera.

Happy holidays everyone.

 

@AcidArrow
Thank you for the reminder why you are on my ignore list...

 

I got impression you pick intestinally points, to suit your cause, but missing my other points. Including scaling down, or writing performant code.

If you can run now 1mln entities on average pc consumer (multiple examples past two years) , like crowd control with formations, giving more cpu cores like servers have, you can easily create complex AI behaviours already. No need 20 years, nor even 5.

Just need people who know how to work on low level optimisation.

Question is, do any game even need that to be interesting and fun to play, beyond just testing for fun. Making it is like procedurally generated open worlds, which can feel empty and boring. So it need have some peruse beyond that.

So realistically more than feasibility of 1mln, we will see such systems with 1-10 or even 100k AI controlled npcs.

 

You are very welcome and also thank you, I don't want you reading my posts either.

 

You can't hang around the GI forum much. That part of the forum would be pretty empty if you ignore acid.

 

This is not Unity in the next 5 years related, anyways the issue of emergent behavior in unexplored territory is one could get "emergent bugs"... well they are not common bugs still they emerge from the laws of the simulation and are unwanted. I'm sure fixing them is time-consuming but that's the matter of emergent simulations.

Has the Dwarf Fortress team fixed the carps bug? Are you a DF player?

 

This is ironic



Small UI align changes and they call it quality of life improvements.

Quality of life update would be if the UV2 generator wasn't complete garbage or if lightmap island packing wasn't complete garbage. Unity have lost touch with reality

 

That's what happens when someone is trying to prove they actually work. Many of these pseudo-QoL improvements sound like actual regressions.

The entire toolkit for baking lightmaps is still unusable for real production and anything related to UX reeks of hostility toward the user.

So nothing really changes.

 

In fairness, UI tightening and improvement of any kind is welcome, I would be excited to see those kind of changes as part of a robust changelog related to lightmapping. The bad part is that these changes are all there is.

 

In 5 years, someone else might have a go at grabbing the steering wheel. Current CEO is getting in on the years. Could be better or far worse for the engine, depends on who it ends up being.

 

The reason why procedural worlds look empty and boring is because they're generated using simple algorithms. Those create patterns and patterns become recognizable.

That's the reason why Dwarf Fortress simulates the process. The game creates t he world and then runs history simulation on it. Cities grow, kingdoms go to war, some beast become legendary, and so on. That results in a more believable experience. On other hand that makes simulation expensive, as generating world takes several minutes, and can take much more if you increase the world size and history length.

 

Shame that Unity disbanded the QoL team that was originally lead by Aras P (before he left). They were doing amazing work!

 

Well, that's sad, that team was one of the few that were delivering high-quality upgrades all the time.

But I don't think they would fix the most infuriating issues. The team responsible for baking either never had resources or wasn't interested in improving the toolkit with modern features/ QoLs required for creating big scenes. The tool feels like designed for tiny linear games made for Xbox 360 / ps3.

But hey, unity bought Ziva and Weta.

 

I look forward to the day us users outside of Unity can actually use these tools. Hopefully in the next 5 years.

 

Probably owned by someone else entirely. Too many other players eating their lunch now, unless a service-oriented revenue model works out. If history is any indicator, I can’t see that happening as UT has a lengthy track record of attention deficit disorder when it comes to developing features and services. I’d be happy to eat those words, but since the 4.x days it has been like this.

I’m in the minority but honestly wouldn’t care about them breaking backwards compatibility (within reason) in favor of quicker development cycles. That is opinion and blasphemy to some.

Somewhere in this post I got sidetracked.. sorry.

 

yeah i know you are joking but my (fantasy) point is, imagine one day in 5 years, UE or any other AAA company releases a game with AAA realistic graphic and amazing quality and in the same year unity release a version where you can get those same graphics with at least 90% of that quality (from the game i described before) running on a low/mid-end hardware or a mobile with less work you do now, at 60 fps minimum

 

The current version of DOTS has brought unity a little bit closer to UE5.
The subscenes approach and workflow is quite good! On converting gameobjects to subscenes, the objects gets converted to entities and uses BRG, thus reducing ton of batches and performance without writing a single line of code! Creating a streaming system on top of subscenes is also quite easy, u can use the one used in megacity demo. Comparing this to unreal 5, their world partition system uses some kind of minimap system which, according to me is better even if it is not best suited for all scenarios, like space exploration games. The unreal system also has out of the box streaming system and HLOD integration which DOTS lacks and u need to author the HLOD's manually.
Even though subscenes is great and easy to use, but the other parts of the DOTS technology aren't, because currently they are very difficult to use, as ECS tech requires u to have some deeper understanding of concepts of C#, C++ and data oriented programming if u want to make game with pure ECS. Currently dots also doesn't support animation, navmesh, terrain, apv. The solution to these must be the hybrid approach between entities and gameobjects which i am trying to figure out.

 

I personally see them as a final nail in the coffin. Unity wasted over a billion to acquire these tools, now they need lots of money to maintain them and potentially connect them with Unity.

But to who? Weta's tools seem to work as packages for Maya. I think it was bought for metaverse, which could never succeed. Ziva, on the other hand, targets AAA devs.

 

My post was half sarcastic and half hopeful.

I also hope that there was some kind of plan in place to how these external tools would be distributed publicly before the $1.6 billion purchase was made.

 

Not to be The Eternal Downer but to be The Eternal Downer for a second...

When was the last time we really saw Unity with a plan for anything like this? I'm genuinely struggling to think when there was an acquisition where it seemed there was a clear plan for further integration or improvement. The Speedtree acquisition seems to be a total wash, for instance. Really the only thing I can think of is Pixyz, which Unity was already pretty heavily involved with even before then.

 

I'm still salty two and a half years after Bolt acquisition. They haven't delivered on any of the promises besides shipping Bolt 1, now rebranded to Visual Scripting with the engine as a package. Two major new features are a buggy new Input System support and wait for it... sticky notes. All they promised years ago will take another two to three years to accomplish.

Basically, everything's on a 5+ year development timeline at Unity. And that's for tools that already lived in the Unity ecosystem. I wouldn't expect Weta tools to surface anytime soon. And aren't most of them Maya plugins, anyway? I'm not sure how/if Unity figures into all of that besides supporting exports better. It's more likely that the toolset will end up being an independent subscription service aimed at 3D artists.

 

The announcement blogpost of the Weta acquisition says:

What it realistically means to be 'connected with Unity platform' I don't think we (or even Unity themselves) know exactly. I read that as having some kind of integrated pipeline with Unity Editor/Engine(?). I guess we have to wait and see. They showed something in that nice-looking uninteractive 20 seconds linear-cinematic small diorama of 2 lions but im curious how it actually could be used for what 99% of people use Unity for.

 

for me, you guys make me feels like they do not have enough devs for everything they want to make/develop and after the bad news of the past year i think it will be worst :/ i always try to be in the positive side but this forum does not help -.-

 

Some folks have a hard time getting off the FUD choo choo train. Keep being a positive voice. Disgruntlement is not a good place to work from. For example.. "The WETA acquisition is doomed". Their plugins are all Maya. Yeah..and MelScript has damn near the same 3D API as Unity..so those plugins are a code translation away from integration. Seems they would have to write a bridge for the mesh and rendering API"s. Mind you there is an entire ecosystem of them so it ain't gonna happen overnite. As a denizen of these forums for over a decade I can confidently say this Unity is doomed banter is a monthly occurrence by those who have a propensity for not completing much at all. Unity makes a great whipping post for them to excuse their unproductive behavior internally.

 

They acquired Bolt with the intention to integrate Bolt 2, a new major version of the tool that took two years of development while Bolt 1 remained in maintenance mode. Unity released one or two beta version of Bolt 2, then cancelled it entirely and integrated the outdated Bolt 1. Before that they worked on DOTS Visual Scripting which was cancelled as well. Turns out that basing an entire Visual Scripting tool on unstable, experimental tech is not a great idea and initially they wanted to have a pure DOTS runtime, which failed and DOTS VS with it. Before DOTS VS they had one publicly announced Monobehaviour based Visual Scripting tool that they cancelled in favor of developing DOTS VS. And before that they had multiple internal prototypes that never saw the light of day even as announcements.

That's around at least 6-8 years of Unity being unable to deliver viable Visual Scripting in Unity. The problem is not lack of engineers, it's management ping ponging all over the place. It's management not understanding the state of Bolt 2 at the time of acquisition. It's management cancelling Monobehaviour based VS tech in favor of banking in on at the time incredibly experimental DOTS/ECS which also didn't pan out in Visual Scripting context. Unity does not lack in engineers, Unity's management lack a strong, long term vision for anything really.