how would you feel if AI generated levels reached human-level quality?

[RHhe82]

12 minutes ago, HotCoffee said:

A.Is will be able to generate content infinite times better than what a human could make and in just a few milliseconds, we will be left with no purpose in life besides consuming something a computer generated for us and we won't have any reason to create anything.

 

In the spirit of Infinite Jest, maybe AI would finally make a piece of perfect entertainment (would it be perfect porn?) that anyone who catches even a glimpse of it would never see any reason to leave the armchair. With a the perfect megawad, only a small slice of humanity would wither away.

Edited September 9, 2022 by RHhe82

[baja blast rd.]

This would be fantastic. 

 

Human plus AI "cyborg" projects -- meaning we collaborate with the AI, using its work as a base -- would still outpace AI-only mapsets by miles. Creativity and ingenuity would probably shift towards other facets that the AI isn't quite good enough at. The humans' role would shift from being core designer to being more of a polisher, remaker, and hypercurator of what the AI creates. So long as the AI isn't superintelligent to the point of making human involvement at all redundant, you could do things like: 

 

- take 20-30 "good" AI maps and add way more of a cohesive narrative to the maps and to the flow of the "megawad" 

(for example even if an AI could make maps that were as good as mouldy's in terms of architecture, construction, encounters -- would the story still unfold in a magical way without some human creation and direction? probably not!) 

 

- sift through 300-400 very good AI maps, find the very best one somehow, and then tune it up even more, substantially remaking aspects of it, hopefully making a "perfect" map in the process  

 

These days everyone is rightfully happy with maps that are by human standards pretty good. The AI would pump those out immediately, and the bar and the artform as a whole would shift towards making maps and mapsets that would now be considered transcendent, but there would still be heavy human involvement required to get from "pretty good" to that level -- again, so long as the AI isn't so good as to render human involvement redundant (although to be honest transcendent maps at the click of a button seems good too).  

 

The artform itself could shift, likely. For example, these days, human mappers have to heavily prioritize what a megawad focuses on, because human energy is limited. But with a good AI, you'd "assign" it one part and do the other parts. It's kind of inconceivable to have something like Stardate 20x7...but it's also a story-oriented wad that has an even richer narrative than Going Down's, and oh yeah it's also better at grand architecture than Mech/Bridge/Gazebo/etc.. Or BTSX...but it's also the definitive exploration of the types of boats you can make in vanilla (the B is for Boat), with every map having a boat that would be #1 on Jimmy's boatlist. Or Counterattack, but it's 32 maps (not all are huge!) and each is set on its own planet or moon and explores a completely new weird-ass gameplay identity. What if Moby Dick or Blood Meridian were a Doom wad, every chapter mapped to a level (and not in the abstract theme way -- you could even recognize many of the lines)? YA novel + Doomkid DeHackEd crossover? I don't know what the fuck that means but I'd love to see it and the AI would answer that. Or on the flipside, since those are all grand and ambitious, you can just take one small map the AI gens that is an architectural shell that really speaks to you, and spend all your time outfitting it with Doomcute, saturating it with atmosphere, tuning the encounters to be just right -- and make the best 10-20 minute experience ever. One gameplay-oriented mapper could spend all their time just encounter designing and reshaping and restructuring areas, and maybe make something that blows current hardwads out of the water, while also looking good thanks to AI.

 

Since an AI could create a "base" to work on very quickly, the AI involvement would be setting down the base -- the 8.5/10-quality maps that you'd pick and choose from out of hundreds and thousands of possibilities, aiming to pick stuff that would be great to build upon -- and the human involvement would be the parts that get it to 10/10+ and also give it a strong human identity. 

 

I think the idea that such an AI existing would make wads soulless is only true of wads that are just pumped out of the box, and the idea that it would make human effort irrelevant is actually backwards! Human effort would be even more important. I'd wager that "cyborg" wads at their best would have even more soul and spirit to them that is currently practical. 

[Kyka]

Did someone say "Hypothetically there would be thousands of AI generated, high-quality megawads appearing everyday for me to play?"

 

I for one welcome our new robot overlords...

[Sena]

One more way to distract myself from playing all the must-play WADs that I should have finished years ago. The one aspect I'd be most interested in seeing would be how the AI comes up with some inherent giveaway to its levels in the same vein as Jimmytexturing.

[DevilMyEyes]

51 minutes ago, HotCoffee said:

I think this kind of problem with A.I will inevitably happen to every type of creative work (drawing, sculpting, painting....) at one point or another.

A.Is will be able to generate content infinite times better than what a human could make and in just a few milliseconds, we will be left with no purpose in life besides consuming something a computer generated for us and we won't have any reason to create anything.

That's a pretty depressing future for us if I gotta say, if it's just generated there's no real effort anymore

Edited September 9, 2022 by DevilMyEyes

[Sena]

38 minutes ago, HotCoffee said:

I think this kind of problem with A.I will inevitably happen to every type of creative work (drawing, sculpting, painting....) at one point or another.

A.Is will be able to generate content infinite times better than what a human could make and in just a few milliseconds, we will be left with no purpose in life besides consuming something a computer generated for us and we won't have any reason to create anything.

 

At risk of derailing the thread, this comment seems to suggest that the only reason people indulge in creativity is because they want to make the best work to have ever been created, when that's really not true. Some people might like to strive for that, but humanity has been around long enough that if this were the case, there really wouldn't be any point in doing anything already - why should I learn how to paint when I know I'll never be as talented as da Vinci?

[yakfak]

i'd stop playing doom and i see no substance in any of the optimistic takes

[SleepyVelvet]

I wouldn't treat it as a depressing existential crisis.  Humans will find what they find interesting interesting, and that will change time-by-time.  There's always some novelty somewhere.  There's always some loose thread in the fabric to pull at, follow and unravel.  A certain amount of hipster-dom is always a part of the human ecosystem/immune-system anyways.

 

---

 

-"Cool, infinite deepfaked Skillsaw maps from cool new bot, Skillharvester"
-"Meh, that got old quick... Oh, a 'My First Map' posted by IAmHumanBean69, a fledgling human and not a matured bot.  Cool, let's check that out"
-"Oh, another 'My First Map' by IAmTotallyHumanBean70..."
-news gets out about AI slipping into the My First Map demand vacuum with dedicated My-First-Map bots, and maybe this third new user, IAreTotallyHumanBean71, isn't actually human...
-"Meh, I'm getting a little bored of My-First-Maps anyways for the time being.  Maybe AI and authentication systems need to sort that out too and maybe I'll circle back"

-"Oh, look at this new cool thing though..."
and then the circle of life continues, ebbs and flows, as it always has...

 

---

 

Doom maps aren't too crazy.  They're mostly a pile of discrete cylindrical sectors that mostly connect to other adjacent sectors in a tree-searchable fashion (shared linedefs, tags, etc).  They be characterized, searched, and smashed into a Deep RNN, or generated from one.  If you make a bot like this call it something edgy too, like Skillharvester, to spook the people who can't see other adjacent existential crisis a mile away. 

 

1 year later from now:  Oh wait, why are there pitchforks and torches outside my window...

[PKr]

That would be a pretty sad day. For AI. I mean look, you can create infinite amount of high quality wads, and then there are only so many humans that could enjoy those wads. Millions of millions wads, and nobody plays them. You get some feedback from optimization algorithm, but that's about it. Pretty depressing. :(

[LouigiVerona]

4 hours ago, Majin said:

in this hypothetical scenario, jimmy neutron the megalomaniacal mastermind creates a program that makes maps indistinguishable from any map made on here. 

 

what do you think will happen to the mapping community?

is there a meaningful difference between AI generation and manual work? 

how would you feel if they were many times better than your own maps in every metric? 

 

Some of the best Doom I'm playing these days are Oblige levels. Modern Oblige is an extremely powerful tool. I have a huge collection of Oblige generated maps some of which I am replaying many-many time for years already.

 

So, we're already there. And I'm loving it.

[magicsofa]

1 hour ago, Biodegradable said:

AI can only imitate and replicate the information we give it.

 

Not exactly... AI can receive and act upon information it has never seen before. It's the logic circuits, i.e. programming, that has for the most part been "static." Getting past that is the key to unleashing a global machine-population problem on ourselves. It's the main thing in terms of "thinking" that obviously separates us from our computers. Humans receive input and react in ways that are often rather predictable, making us appear machine-like. But we do have the ability to intentionally change the pathways in our own brains, so that our reaction to the same stimuli can diverge.

This is why machine learning is a different, one-layer-more-complex process than conventional programming. Before it was X in, Y out, and the same X would always give the same Y. Now people have succeeded in making programs that are capable of reflecting on the result Y and potentially deciding to change its own logic based on whether the result was desirable or not. It's still got a human telling it what is desirable and what is not, but definitely more dynamic than your average computer program. And besides, don't humans have a DNA blueprint, which grows into default systems that already have many answers built in? You didn't just start from scratch - you were in a sense "programmed" from the beginning, just with a high level of dynamic ability allowing you to learn quite a lot during your life. 

[Naarok0fkor]

I'm already jealous of Midjourney...

[Faceman2000]

I don't know about all the rest of you, but there is already way too much content for me to keep up with. More content would not inherently affect my life in any meaningful way.

 

Similarly, the existence of "perfect" AI-produced levels wouldn't change any of the reasons that I map and make levels for Doom - as a way to express creativity and realize the things I've envisioned - and so I would keep mapping. I feel this is true for most people who map.

 

In the end, nothing would really change all that much.

[Kinsie]

4 hours ago, Majin said:

in this hypothetical scenario, jimmy neutron the megalomaniacal mastermind creates a program that makes maps indistinguishable from any map made on here. 

 

what do you think will happen to the mapping community?

The big thing that limits AI as we see it nowadays from stuff like DALL-E etc. is that its very nature of being trained off of pre-existing content means that it can only work with old ideas. Humans still have the upper-hand, in that they can come up with exciting new ideas. The robots might be able to whack some cool sectors together, but they probably can't come up with the idea of wrapping it with a whole new conceit like, say, going on a peyote-fuelled spirit journey to Mars.

 

There are also potentially interesting ideas in using AI to find inspiration. Arvy, the author of Diabolus Ex, did some challenges a while back where he told the Craiyon image generator to create smeary ghosts of videogame user interfaces, which he then recreated into actual things.

 

 

4 hours ago, Majin said:

is there a meaningful difference between AI generation and manual work? 

To shamelessly steal/paraphrase a line from the aformentioned Arvy, it's like the difference between an artist and an art director. One is drawing stuff, one is telling someone/something what and how to draw stuff. Entirely different skillsets and practices.

 

4 hours ago, Majin said:

how would you feel if they were many times better than your own maps in every metric? 

OBLIGE/OBSIDIAN is already a lot better at mapping than I am, so not much change there!

[Deadwing]

TBH, I wouldn't be able to distinguish Oblige-generated maps with "humans" ones, unless I played a lot of them to find the patterns. Some of my older maps were even described as oblige-like (not in a negative way, also, afair).

 

Funnily though, I do enjoy more playing (mostly) one-man megawads due to how the mapper handles these patterns and find new creative solutions to them, creating a large sense of "narrative", if that even makes sense 

[TheMagicMushroomMan]

We'd need a Robo-Doomworld to go along with it. AI mods, members, maps, AI debates and discussions. I'd read it everyday just for the AI drama, the hell with the maps.

 

Oh, and they still have Post Hell on Robo-Doomworld, for all the naughty Roombas.

Edited September 9, 2022 by TheMagicMushroomMan

[Solmyr]

3 hours ago, map11has2names said:

holy shit. that could really be the end.

It doesn't have to be, the community could still pull through and survive even in a worst case scenario (90% of userbase quits and wadhosting sites close), and after other members pointed out that Slige/Oblige/Obsidian exist and those generate good maps/megawads (especially Obsidian), we're halfway through to advanced AI super-high quality megawads, people didn't stop making wads so far and the community is still thriving.

 

I guess there's no point in resisting change as it's inevitable.

Edited September 9, 2022 by Solmyr

[CyberosLeopard]

Personally I’m a computer nerd (‘though not a very good one), so I would honestly love to see an expert system generate maps that are indistinguishable from the best mappers the community has to offer.

 

But like Faceman2000, this wouldn’t change anything for me. I have not played most of the acclaimed maps and map-sets, and I don’t work on my Doom projects with the purpose of trying to keep pace with that of others (and I really need to tell myself that more often…).

[baja blast rd.]

1 hour ago, Faceman2000 said:

I don't know about all the rest of you, but there is already way too much content for me to keep up with. More content would not inherently affect my life in any meaningful way.

 

Similarly, the existence of "perfect" AI-produced levels wouldn't change any of the reasons that I map and make levels for Doom - as a way to express creativity and realize the things I've envisioned - and so I would keep mapping. I feel this is true for most people who map.

 

In the end, nothing would really change all that much.

 

I think if there was an influx of tons and tons of 9.0/10+ AI-gen maps, with that being ubiquitous, the bar would just raise to people only wanting to play 9.9/10+ stuff that is generated this way. And usually very high-effort human-AI hybrid wads. The bar would also not be only about quality, it would also be about finding niches that you like (arguably more important anyway imo), like "I'll only play such a wad if it's earth-shatteringly good and it uses any of the highly specific experiential strategies that I want in a wad this month." If you set the bar to only involve a quality threshold, that gets overwhelming in a hurry. But niches would narrow it down just as fast. For example: "I'll play this 9.9/10+ AI+human wad with a highly idiosyncratic premise and story, with difficulty around Speed of Doom, with open-ended combat (minimal setpieces) and odd gimmicky fights and 'biome-oriented design' (aka when the setting within a map largely influences the combat that is found within it, e.g. all slaughter is in one or two types of setting within a map, another setting might be always somewhat generous with ammo, another setting might always have lots of airborne monsters, etc.) that also does the 3iaC thing where the map has three 'protagonists' with their favorite quirks and weapon loadouts. But I won't play this other 9.9+/10 AI+human wad because I'm not in the mood for something that is set in restaurants and tasks you to fight monsters to find the ingredients within the map to serve an end-level cyberdemon and, depending on the quality of the meal, the cybie is either happy or gets upset and fights you in the cramped restaurant. I'll save that one for next month!"  

 

It feels like there is "too much content for people to keep up with" now because people tend to be relatively broad in what they play, and selection boils down to more like, "I can play any of the good and interesting wads that aren't in niches that are deal-breakers for me." 

 

But if people got more selective, and curation styles were more geared towards listing out specific experiential priorities and tropes and all of that, that might be less of an issue. (If someone poured a lot of effort into doing that now, then people would realize that while they might  potentially enjoy way more wads than they can ever realistically play, a much smaller subset of those would be super high-priority can't-miss experiences for them.)

 

In such a world, being "hooked" by a wad would be a lot more like being hooked by a novel. These days the wad pitch is: great screenshots, great concept (1 sentence or something), author whose work we love, playing a bit of it and being hooked by gameplay, seeing other people play it and realizing it would be fun -- stuff like that. if ultra-HQ wads were ubiquitous, pitches would be way more "specific," in a way (sort of like the descriptions of imaginary wads I used above). And the cool thing about pitches like that is that even if you had 100-200 "perfect" wads to possibly play, some of them would clearly stand apart as way more appealing based on their pitch. Wads that are defined exclusively by being "really good" at general attributes would fall by the wayside. "Really goodness" would be the default, so it'd be almost exclusively about the creativity the human collaborator could infuse/remake into the work.

 

It also feels like a not-insignificant part of the value of experiencing a wad is in knowing that other people played it and had opinions on it. Experiencing art is partly a communal thing. It would be lonely to play an amazing wad that no one else ever played or ever will play. So maybe more of people's playing would pivot to DWMC-like activities where you could be assured that many people are playing and talking about the same wad that you are. These days if we boot up a good wad, we know many other people will have played and enjoyed it too; reviews and Discourse(TM) would exist around it. If there were 150 10/10 wads a year, that would not be assured by itself for most wads. So DWMC-like clubs would give structure to that. (There would be people who risk it, though, and venture into wads no one else has played -- and then report to us that they are really worth it.)

 

(Also I think very strongly that more "primitive" human wads would still have their appeal. Sort of the way that I have played lots of classics and all-timers but I still open up 5-minute My First Wads -- and enjoy them non-ironically. You can even argue that these wads could still exceed soulful AI-human hybrids by being even more soulful. Imo 'quality' is probably a bit overrated in itself.) 

[DannyMan]

Doom's Knight 2 would be possible then

[RastaManGames]

"Go ahead, big boy AI, and generate me an entire megawad on city/urban theme with unique levels, styles and places..."

[Sectorslayer]

Just add AIs to post threads on Doomworld, playtest and upload to IDGames until it is done. 👹

[D0M0]

I would just use them for completely stupid experiments, maybe one day i could get a full DIY workshop

[indigotyrian]

I think people have a tendency to assign way too much agency to artificial intelligence in general. AI transformers (midjourney, dall-e, etc etc), just like all other forms of AI, are tools; they exist to extend and expand the capabilities of raw human performance. However, they can't replace the human at the center. There have been freakouts about tools replacing humans for a very long time and those fears always subside as said tools end up uplifting countless people.

 

I imagine a sufficiently advanced AI that can automate a lot of the "busywork" and heavy lifting that's associated with mapping would lead to an explosion of creativity like what baja blast described.

[Savarin]

Simply put, I would be pretty happy. I got tons of new work to get inspired by, and now I can output work much faster, thanks to the AI doing the dirty, repetitive work for me. Imagine a mapping AI where you can toy with pre-built themes, stats and all kinds of map variables, maybe even create your own presets!

 

I view it the same as animation Interpolation. Yes, you could put less effort into animating your frames, doing the least you can and letting the program do everything else... OR you could use it as a tool and work faster. I could lazily generate Megawads everyday that look indistinguishable from eachother, but who wants that? Instead you could work out a fuckin AWESOME Megawad that would take 1 year at least, but in a month instead... or rather, let the tool do the boring part while you learn to do cooler and cooler stuff with mapping.

[Faceman2000]

Maybe Mordeth could be finished by an AI.

[BigBoy91]

Shit would be tight.

[Andromeda]

8 hours ago, Biodegradable said:

20 years ago when @fraggle set an old version of SLIGE to fart out a new megaWAD every single day that's presumably still functioning to this very day(?) and yet, here you all are, 20 years later, still crafting maps.

I wonder if all these wads have been archived somewhere...

[Alex S.]

One thing AI cannot replicate is the true love and passion of a dedicated creator. I would not feel like there was the same level of quality in anything AI-generated because a true fan and creator must be a human being in order to put out a good map.

 

[Captain POLAND]

What if they started randomly including Terry WADs in every batch of WADs they designed?