Jump to content
The Dark Mod Forums

demagogue

Development Role
  • Posts

    5818
  • Joined

  • Last visited

  • Days Won

    82

demagogue last won the day on April 25

demagogue had the most liked content!

Reputation

1363 Deity

About demagogue

Profile Information

  • Gender
    Male
  • Location
    Tokyo
  • Interests
    international law, cognitive science, piano/guitar

Recent Profile Visitors

16850 profile views
  1. Just in terms of the raw tech, human brains still have ~8 orders of magnitude more nodes or multiplications (10K vs 1Q synapses I think it was), even assuming you equated nodes between neural nets, which is misleading, but it still handwaves at the distance artificial neural nets lag behind in pure information terms. We're still in earthworm territory and need high end graphics cards to crunch that much. That's not a mysterian argument, but it has a similar punchline. When your model lags that far behind, it may as well be magic what the brain can do in comparison. It's interesting that you can do (what seem to us like) really high level things, like this procedural art, chess engines, and probably music and other forms soon, with an earthworm sized brain; but really (what seem to us like) simple things, like ordering lunch at a fast food joint or other open ended things, run into the AI Complete problem, i.e., you can't even do the most simple operation unless you have full human-level cognition and world knowledge.
  2. From what I've been reading in the AI lit, the next step is rapidly integrating new information into the network weights with some tricks to avoid the so-called catastrophic forgetting problem, more in the Grossbergian version of neural nets to get a little technical about it, more in the way human memory actually works. Humans also don't save visual images verbatim either, but they're procedurally recreated in memory on demand. Basically what it means is you can get your local AI to look at a work of art or music and it'll recalibrate its weights so it can "recall" it later based on some lightweight cue like "that painting I uploaded on (date & time)". I mean the user doesn't even have to know it's not a direct save and it got embedded in AI weights. ------- I remember in high school and college reading the claptrap from some futurists about the coming Singularity, and laughing it off, not even whether we were actually approaching it, but that the very idea itself felt kind of too outlandish to buy. But now I'm starting to get a sense of what it might actually be like to reach that point, when AI have integrated knowledge way beyond what humans can follow, and humans having direct and instant access to it, they just take it for granted like it was an extension of themselves. Or something like that. I'm still not sure how it may play out if at all.
  3. I can imagine when this tech goes into the third dimension it'll be a big sea change. I can even imagine in the future the tech procedurally generating the gameplay along with it, so you can just generate a whole game world, characters, & gameplay. I don't know if the tech for storytelling will be up to scratch on the large scale, it's lagging so far, but on the small scale I could see it convincing enough. ----- What's been striking me most is that there's just such a flood of it. I have my folders of AI art... Actually I started them like 10 years ago, but I only ever collected a few things more for the gimmick of it than it actually looking good. But now I've already got masses of really interesting works in them and find masses more every day. But I think there's a point where there's just too much. That's what I think is going to be the big issue on the social and economic level. There's so much that one can only take so much in, and what does that do for anyone or anything else that wants to squeeze into that space.
  4. Y'all should visit the MidJourney discord channel if you haven't already. The other big thing, the FOSS one, is Stable Diffusion. That's what I've been using (via the hugging space demo, which doesn't use a credit system), and it's typically awesome once you know what you're doing. But I don't know where people are posting it in a single place. It's coming fast and thick now. I'd say half of it is awesome, and the other near half is awesome but there's some little screwy bit you could fix up in a half hour. Then maybe 10% are duds. Of course the posts are going to be biased on the awesome end just because they've been filtered. The other thing I've noticed is suddenly a ton of art on Twitter and FB is AI art, and again it's typically great. Yeah, art belongs to the computers now. You can be in denial about it, and it'll take the next year or two to really settle in. But I think the die has been cast now.
  5. There's always two relevant versions of the source code, the version for the latest release and the WIP version. If you're in it for learning early on, if you want to do small little experiments where you just swap new source code in and out without changing assets, or if you have space issues or want to keep things compact, or if it's on some really niche or independent part of the game that updates probably won't touch it, etc., then the release version may be fine for your purposes. But if you want to get into it seriously, at some point you'll want the WIP version as an independent install, anyway, because it's kind of self-defeating to work on code that's already been made incompatible, and once you do it, it frees you up to do lots of different things without worrying about what it's doing to the game you also want to play sometimes. As an aside, I really encourage people to play with the source code. You learn a lot about coding generally, since it's always best to learn by doing, and by doing something you're interested in instead of boring textbook problems. And the kinds of experiments you can run on TDM tend to be fun because, as an immersive sim, there are just a lot of fun things you can do with the game and the world. I think a good way to start, after reading through the code a bit to get a feel for where things are, is to give yourself some random small assignments, like putting in little features or gimmicks, and then search the code for the elements that seem like they'd be most relevant for that thing. And then as you start tweaking things, the problems you run into are going to be teaching you how things fit together. It's a fun way to learn though.
  6. I've been using Stable Diffusion a lot recently. Something you have to note is the current way to use it, you should expect to make 20~30 versions. One of them is going to be awesome. The other thing is that you often have to re-do wonky parts, which most of these systems are starting to do. I mean maybe 60% of it will be right on, so you keep that and fix up the wonky bits until they fit. So you're not getting perfection on the first run. You're iterating the same image in bits in pieces, each block taking 20 iterations or so. Someone was saying it before .... hands, feet, legs, eyes, will look a little wonky. So you just drag a box over it and start iterating, and eventually it will get it right. By the end of that you're going to start getting close to perfection. We have the tech for that already now. It may take a little time, but you can get perfection with what we have now if you block iterate it enough. And the final thing is, look at what Dall-E was making just 1~2 years ago. It was trash, and what we have now is lightyears already beyond it. In another 3~5 years I think it'll meet the hype, and every thing churning out of it will be photo real or whatever you're looking for, and then some. I don't think it'll take that long to get there just looking at the differential slope of the progress we're seeing right where we are now. I think of it like chess engines. There was that period where people were wondering if it'd beat humans, and then it did, and then it just kept going without looking backwards until at this point the entire chess world looks like it's going to run into a crisis of conscience because the game is out of human hands now. There's every reason to expect imaging software is going to do the same thing in short order. For gaming, e.g., I'm making a card game, and this is helping me make the background and card images. Again it's taking like 20~30 iterations, and then I have to fix a few things by hand, but I can churn out a half dozen really good looking cards in a day where I wouldn't have been able to make a single one before. But like I said, I expect in 5 years I won't even need many iterations or the hand-fixing... I think the fixing will be automated well enough. Oh another thing is, because it's still diffusion over a data set, the big limits are users not understanding the ontology embedded in the set. If you figure out good search terms, you can get what you want much better. Another thing I think will improve in the future isn't only the tech, but the search terms and other UI elements will work much better to make the system give you quality work that you want, by improving the data set, but also you can have a second-order of AI that parses text inputs into a form that works with the data set. I think that's part of the next generation of AI that's going to make it start much better delivering to us on a human level. I mean I think the tech is already there now in principle. Now it's a matter of making the tech work for humans, and that'll be the next big revolution coming soon.
  7. Does anyone ever actually read these status updates? 🤔

    1. Show previous comments  11 more
    2. datiswous

      datiswous

      To sum it up: Everyone under your status update (including you) reads (your) status update(s).

    3. Dragofer

      Dragofer

      Yes, but I doubt everyone who reads the status update replies to it.

      Also, sometimes people reply to something without having read it.

    4. AluminumHaste
  8. I was busy before and missed this (the Seeking Lady Leicester WIP screenshots), so I guess it's good this thread was bumped. It looks really great! I'm excited to play it! I'm really happy there is this outlet for us to see part of grayman's great legacy still emerging in a way that respects and even fulfills his wishes. And I'm happy we have reached this approach with Biker for him to contribute to the community in a way that's safe, controlled, and ultimately healthier for Biker & the forum. It's working great. Don't change a thing. Above all I'm happy we have great FMs still being worked on and coming out. They really are the best medicine when everything else in the world seems to be falling apart. The SEED system was created exactly for this purpose, it looks great when it's used, very realistic, and it should be used more often!
  9. Edit: Oh, I guess I thought I was on the last page when I posted this and I wasn't. Well anyway, since we can't delete posts, here's a reply to a probably year's old post now... For my FM I mentioned a key window being openable in a readable, and I put a loading screen message that some windows were openable. I think I also put a slight light in front of the openable windows, or some of them, so they popped out. I wasn't such a fan at the time of the meta-diegetic route (some with knobs and some without), and leaving one open seemed counter-productive for performance reasons, but performance was a much bigger deal when I made it than it is now.
  10. Oh nosound may be a sound shader too. I was working with "silence" the sound file & associated sound shader for my own project, and the function nosound in the code itself (where it means what I said). But if you saw nosound as a soundshader, then I'm sure you were right in what you saw, and if you want to know the difference, you could open up the soundshaders and see. Also note that I worked on this stuff ages ago, so you don't have to trust everything I'm saying too much. I just get a kick when people talk about code I worked on and I want to talk about it. Yeah the stereo issue came up later, and I'm glad somebody took care of it!
  11. To be technical about it, "nosound" is a script function to stop a speaker. "Silence" is an actual sound .wav file that plays nothing, so you can still have the speaker "play" an arbitrary .wav file, e.g., for the ambient system, then use that to easily not get a sound, as you might do in the popping case as mentioned... Or that was the theory. (The popping problem was fixed after I added that, but I was never 100% sure that was the fix that did it.) But the issue in any event now is that now countless maps already use "silence" references in their location_settings spawnargs, and it's dicey changing things that mappers add themselves because they won't update with TDM changes, e.g, as opposed to using stock resources that automatically carry TDM changes with them into new versions.
  12. If you'll allow me to take a bird's eye view of this ... remember back in 2016, when they nutcakes were campaigning for Brexit in the UK and Trump in the US? The joke back then was, no matter how crazy one side got, the other had to outdo it. And I was wondering which flavor of crazy was going to be more permanently damaging to their respective country. At the time I thought it was going to be Brexit, since that's "forever" and Trump was only 4 or 8 years, but I underestimated how deeply Trump's admin could undermine all of our institutions., even after he's left. But then I also underestimated the wreckage that was going to be left in Brexit's wake. ------- Fast forward 6 years later and I'm still not sure who has the worse deal. Even if this law doesn't pass, it seems like it's putting on display how dysfunctional UK politics is now, like the ultimate exercise in dissociation -- trying so hard to ignore how shambolic the post-Brexit political & economic situation is by hyperfixating on problems that aren't problems with truly batshit legislation. (You can correct me if my reading of it is off.) But if you've been paying attention to the Jan 6 hearings in the US, you know the US still won't be outdone. The problem isn't that the hearings are bad. The problem is that they're a slam dunk case against the previous administration and GOP, but it's only (apparently) making the whole GOP circle the wagons and double down on their insanity. Not to mention the SCOTUS decision that's going to allow the criminalization of abortions, the arguments behind it could just as easily apply to criminalize birth control, or gay or transgender behavior, and I have to imagine in some states there's a sizeable part of the population chomping on the bit to do just that. ------- On the topic of this law itself, of course it's a mess. It's weirdly chauvinistic. UK-legal-compliant browsers & the extraterritorial bit? Didn't the UK just go to a lot of effort to tell everybody it was done with multilateralism and would be much happier as a fading second rate power? It's a tall order to follow that up with this. But the big problem is the part about net neutrality. The more control the state wrangles the Internet into, the more it's just going to whip it into some propaganda mouthpiece for megalomaniac & tyrannical demagogues (& not the benevolent kind like me). I'm all for strict measures to protect children from child pornography, get it off the net, and punish anybody uploading it or any server or site hosting it. But forcing people to maintain their identity for sites is asking for people to be targeted, tracked, and coerced, or putting in the architecture to do that when the bad guys get into power. The problem is even if this law fails, I think they may keep trying until they make it work. This is all a stupid road to dystopia we're on.
  13. My memory may be failing me, but I thought there was a key you could press so that frob helper turned on while it was pressed. It may have been that somebody gave instructions to bind a key to do that job through the console. But anyway I really liked that system. It would stay off, but when I needed it, I could press that button and have it when I needed and wanted it. (And IIRC this is what I've done, but it was already a few years ago now(?), so I don't remember all that well. I did try it with it always on for a while, reflected on it, and still didn't like it, XD or didn't like it popping up when I wasn't asking for it. But I liked being able to easily turn it on & off. But this is just me; I don't presume to speak for a whole demographic or anything.) And like I said before, the team can do whatever it feels is best as long as I or anyone can still set that up, or set it up however they like.
  14. I was using the term more metaphorically. There was a design philosophy. Whatever the literal doc itself said wasn't really the point. Well I shouldn't have referred to it as such since it wouldn't be on point anyway. Most things got pounded out in the dev forum across countless threads, but that's harder to explain. I was using the term "design doc" as a cheaty shorthand so I wouldn't have to explain. XD Anyway, long story short, if there's a vote among the devs to change a thing, I trust everyone on the team has the best interest of the game in mind, so of course it's proper they can change something like this that way. I wouldn't vote for it, but I'd respect the vote and be fine as long as I could turn it off. I think one characteristic of this as opposed to the light gem or inventory widget is that at the end of the day this is reticle, and reticles carry a lot of baggage with them in games. Somehow they change the relationship of the player to the game space, at least with my feeling of them. I always turn off reticles in any game I play for that reason (although like I said before, I'll cheat sometimes and turn them on in rare cases; but I don't like them in any game as a rule). So that's one part of this.
  15. I was thinking about things like frobbing coins and doors, but you're right about containers being an innovation, and more generally that realism and era don't match up all that well when you get down to details. But that wasn't really part of my point, so it's best to rephrase it. The point is there is a category of player that hates screen bling, or will say they hate it (we're really talking about a whole philosophy towards games & immersive sims), and that type would view TDM is the kind of game where you can get relief from it (etc.). I don't know the best way to summarize that in bumper sticker form for a button. Any way you try there are going to be problems, like you're saying. But it's still a real category. I'm open to suggestions for the best way to succinctly describe what it is. I was trying to throw a bone to people that don't like the default and wanted another option that worked for the other player type that says frob-hassle is a pointless frustration, but I agree with this. Having some big-scale option setting moment is a bad idea, so I'd even retract the suggestion. But that would just lead me back to: ideally we keep it the way it is because it's consistent with the design doc from the very beginning. You can't micromanage or tweak a design philosophy forever; it's a kind of endless feature creep. Now that I think about it, that's probably closer to the root issue. There are always pros and cons with different features. And here if you're talking about two different types of attitude, which the same person can even go back & forth on, you can't win. But at a certain point a feature becomes part of the character of the game, so it's best to stick with it. That's only my view though. (And I like that frob help is an option out there I can use sometimes.) Of course another option is what Springheel likes to do with his loading screens, which is having hint texts on the Main Menu screen that inform players they can do stuff like this that they might never notice otherwise. That's a good idea for other reasons because there are quite a few things that would be hard for players to know but which they may find really useful.
×
×
  • Create New...