27
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 04 Nov 2024
27 points (100.0% liked)
TechTakes
1403 readers
87 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
You know that thing that happened with the AI generated DOOM?
Well, someone decided to do the same thing with Minecraft, and you can see that the results are... basically abysmal:
https://youtu.be/7Jd-Rr9cJYo?si=-9XZ51ss6cBuSiC3 Skip to 2:00 for the actual "gameplay".
TL:DW nothing is saved outside of the view screen, things aren't even saved within the view screen, the resolution is like 240p at 20fps, input latency and mouse latency is awful, and this was all apparently done by training on literal millions of hours of Minecraft footage. The mid-range computer I had from 2006 could run the game better than this. A 14-year-old netbook could run the game better than whatever supercomputer they're using to render this.
Note that the person in the video isn't part of the team/whoever that created it, just someone who is reviewing it.
maybe i'm a weirdo but i actually really like this a lot. if there weren't armies of sycophants chanting outside of all our collective windows about how AI is the future of gaming... if you look at this "game" as an art object unto itself i think it is actually really engaging
it reminds me of other "games" like Marian Kleineberg's Wave Function Collapse and Bananaft's Yedoma Globula. there's one other on the tip of my tongue where you uploaded an image and it constantly reprojected the image onto the walls of a first-person walking simulator, but i don't recall the name
I do like it also, but notice how they try to hide most of these strange effects in their demo video. They're trying to pass it off as a playable minecraft. Anything interesting about the technology is a defect to these people.
They also claim this is "the first playable AI-generated game" when the exact same sort of thing already existed in 2022 (and it looked a lot more efficient and haunted): https://madebyoll.in/posts/game_emulation_via_dnn/ Maybe it doesn't count if you call it "Neural network" instead of "AI".
That article gave me a whiplash. First part: pretty cool. Second part: deeply questionable.
For example these two paragraphs from sections 'problem with code' and 'magic of data':
Well, "just read the dataset bro" sound great sounds great until you are staring at a dataset with 100 000 examples and someone is asking you to interpret it.
What the hell is this even arguing for? Is one module with ten million lines somehow better than 100 modules à 100k lines?
yeah, that "most of the internet will be Al-generated" nonsense is tanking my ability to take them as domain experts seriously.
still, something gets me about completely generated, transient-when-you're-not-looking, constantly shifting worlds. might have to collect more examples
It's dream logic.
Showed this to my wife.
See, there's the mistake. You can't let outside people actually interact with the thing; you need to stick with cherry-picked 2-3s clips.
I've seen this floating about quite a bit, and everyone I know is dunking on it - the most frequent comparison I've seen is calling it "Minecraft with dementia".
It also shows why those DOOM demos were only 2-3 seconds long, because that's how long it can keep cohesion for.
oh cool, Minecraft LSD: Dream Emulator edition
which would be a cool concept, if the generative AI model weren’t incredibly prohibitively expensive to run, trained on plagiarized videos, and incapable of coherently tracking state (believe it or not, LSD: Dream Emulator does have a gameplay and progression loop… game-like things without one tend to get dull very quick)
This has been a thing with games leaning heavily on procedural generated content as well. Takes a lot of effort to make those more interesting in the longer term. The ai is the future of gaming people are sure to rediscover this fact again. Sadly compared ro pgc people they will waste a lot more energy and money.
I think it's fine to have little art games that are only interesting to play for a few minutes. They can still have an impact that stays with you forever. Catacombs of Solaris for instance is one I love and think about all the time.
These AI guys though, god it must be so sad working on this stuff. They can't be proud of what they're making if the goal is to trick people into thinking you can have a real Minecraft game in there. The game is not even relevant, as far as I can tell they're only using it to draw attention to their stupid and doomed ASIC business.
goddammit! you have no idea how many variations of "first person walking simulator projected image texture trippy visuals" i slapped into every search engine!
but yes, that was the one i was thinking of
Ah awesome! I didn't think it was because you can't upload pictures, but maybe you can in the "Revisited" version!
i went and bought it, and yup, the revisited version is the one i was thinking of. time to walk around inside a picture of Sam Altman so i can absorb his raw intellect and business acumen
Pgc can be a useful tool in the toolbox however. Just look at dwarf fortress for example. But games need more than just that. And with LLMs there could also ve some use (generation of voices I think for example, or some low level of conversational chatter). The problem is the data usage, the anti worker stuff (no voice actor wants their voice stolen for example), and like in this case, the people just promoting their other crap. Like how crypto games are just a shell to promote their crypto bs.
LLMs could have some minor use but with the backlash due to techbros going their usual vc style techbro ways, and steam requiring disclosure I doubt it will be of much use the risk is too great. Which considering the power costs is great.
We also have stuff like DLSS, which is pretty cool and I imagine ends up being more efficient in energy usage. The key thing I believe is it's built as a specific tool for a specific purpose, instead of general purpose slop machines eating humongous databases to excrete absolutely nothing useful for anyone.