I'm a 19-year veteran of the games industry (im 40 now), in AAA and indie - worked at Ubisoft, Rare, EA, you name it! The only people clamouring for genAI are, sadly, the top floor executives with zero actual development experience. They see Magic Robot That Creates Anything You Desire and go "ka-ching!," without considering for one second how the model was trained, whose work was stolen to train it, and quite importantly how long this wild west can continue before governments start to litigate on intellectual property ownership rights extensions, compensation for targets of training, and more. Effectively, whilst the technology will improve, that's not to say every country will permit this unimpeded data grab/IP skimming forever. Ultimately I see it like this, bud - Pot Noodle and Spam did not kill off restaurant dining. GenAI fills the niche of the "machine-made slurry of questionable origin which doesn't overly impress nor satisfy, but does the job in a pinch." Like spam, some will see that as good enough to consume every day with no regard beyond a full belly. Others will see it as a last resort - a "when all else fails" option. And others, and I'd argue a fairly decent number, they'd simply rather spend a little extra now and again to sit in a person's restaurant with a loved one, and dine knowing it was hand-crafted with love; to enjoy each carefully-crafted morsel and refer back to them fondly when they do have to eat a Pot Noodle a week before payday. I don't see GenAI as replacing me for the games that I make nor the customers I make my games for. I see GenAI as making for a customer who only sees content, not craft. Who doesn't care if the eggs are free range or not... they just want the cheapest eggs. Those who don't care if Nestle steals water from the poor and needy, because they want a kit kat. Simply put, those who respect beyond the surface level will always want human-made games. We may lose some customers overall... but those were just people who found cheaper Pot noodles and spam.
@bencesarvari2235 You say that like it's a bad thing, but I think that sounds like a super interesting experience. You sold me on this. im going to try this out.
The longer the AI hype cycle continues, the less I'm worried about it when it comes to art, honestly. The comparison to NFTs, I think, is very apt. Something that blows up, grifters looking to make a quick buck (or a lot of bucks) dive all the way into it, and then people catch on that its got no clothes and the market for it crashes to nothing. People talk about how "amazing" AI is, and how good it's gotten recently, but I've never seen a bit of it that didn't suck. "But it can make videos that..." yeah, I saw the woman walking down the street. I saw her jacket randomly morph when some of it was outside the screen for a moment. The fundamental problem with AI, as it stands now, is that AI doesn't know anything. It doesn't know that 2 + 2 = 4. If you ask it for the answer to 2 + 2, it calculates that many other people have said 4 in the past, so it says 4. But it doesn't know that. This may seem like a pedantic distinction, but it's the core of why AI is so bad outside of tiny little windows of use, and why even in those tiny windows it needs to be watched like a hawk so it doesn't do something like tell you to Elmer's your pepperoni to your pizza. I don't think anything has demonstrated this so clearly as AI Minecraft. Just think about what had to happen for MAInecraft to exist: 1. Create Minecraft 2. Minecraft gets insanely popular 3. Minecraft becomes one of the most-streamed games of all time (or is it THE most-streamed game?) 4. Tech company spends probably millions of dollars to train a model on Minecraft videos 5. Tech company releases a broken mess that can't remember anything that happens offscreen because AI doesn't know anything "But it's going to get better!" people like to say. But, let me tell you a secret: it won't. Literally, the LLMs are out of data to hoover up; they're as good as they're going to get. And they're not good. But more importantly, the fact that AI doesn't know anything is the structure of the system. It's not something that can be "fixed". Someone would have to invent an entirely new technology, at which point we're talking about something else. But also, look at what it took to make MAInecraft. Look at step one! The only way to make a "game" with AI... is to have already made that game! Not just already made that game, but it has to be a simple game, with a very consistent UI and look, and millions of hours of available video to suck into the model. And then, the AI can spit out a Cronenbergian nightmare that, in some single frames, looks similar to the original. Yes, they'll reach a point where they can say, "Make Minecraft, except replace the sheep with facehuggers from the Alien game I also trained you on". But that just means they have to make TWO games before they can make their one bad AI game. And the more complex the game, and the more you feed into the model, the less the AI will be able to handle it. Just look at the inventory in MAInecraft. Someday, I'm sure they'll get that into the model in a usable-ish manner (except it won't remember what items you have because it can't know anything). But imagine if the model was trained on a hundred different games, and you tried to "prompt" the UI into a usable system, and "prompt" how inventory worked. The AI can't handle that. And the poor employees that are ordered to make it work will try, I'm sure, and get fired when they can't deliver, I'm sure. But, quite simply, the technology is not able to do it. There are plenty of areas that I'm worried about AI. Surveillance, law enforcement, internet search, driving, propaganda, etc.. Mostly, because it's going to be "lying at scale" and people are going to rely on those lies and die (see: mushroom foraging as written by AI), and/or because it's going to remove accountability from a critical decision (no person crashed the car, the AI crashed the car). Art will be as fine as it ever was. Artists might have a hard time while foolish bosses fire them expecting AI to fill the gaps, but AI will not fill the gaps. Nobody's going to buy MAInecraft, or the movie version of it, or the novel version of it. Because it's bad. Even images, which is arguably where "AI art" is strongest, the result is very samey, very soulless, very obvious. The people wanting it didn't want art in the first place. To take Adam Savage's words and treat them like my own: Art is about presenting a point of view. A machine does not have a point of view, so what it makes is not art.
On your point about creating an entirely new technology, they would still call it AI. This isn't AI either; they call this language model stuff AI for marketing purposes. So if they created an intelligent AI, it would still be called AI, my brother.
@@aldiascholarofthefirstsin1051 They would call it AI. That changes literally nothing about what I said. The point is that the technology to do what people claim LLMs can be "improved" to is not possible with the current technology. Something new, which does not currently exist in any form, will need to be invented. The names of these things is irrelevant; what's important is the technology.
Seen some AI minecraft. Person was climbing a hill and once the hill filled the screen, the AI's lack of object permanence kicked in and decided he was looking directly at the floor. E: Fill screen with sheep. It is now a mountain side.
The developers of Oasis had to forego most of the context buffer to make the experimental project run more smoothly. This thing is really more of a novelty/proof of concept than a game.
It's really impressive that a computer can (sort of) simulate Minecraft without actually playing Minecraft, but what's even the practical application? Nobody wants to play the weird incoherent LSD Dream Emulator version of Minecraft, it probably takes way more tech to actually run than the real thing, it's nauseating to look at, and completely unplayable as a game. So. Who is it for? What purpose does it serve? Why were so many resources put into making it? Who will even care about it? It's like spending billions of dollars and hundreds of thousands of hours of research and labor to develop a 5 hour long process for putting on your shoes.
What I hate about Silicon Valley's promotion of generative AI, is that instead of seeing what it does well with and applying it to something useful and social good, it's just being shoved everywhere hoping to be the first to profit. Billions have been invested and there hasn't been much ROI, they're getting more desperate as time goes on. A great use would be generating NPC banter as you walk through a city. That way it wouldn't repeat and it's not critical path content.
As someone who works as a programmer and did some work with AI in school, I can safely say that I'm not really concerned with AI in the long term. Artists and creatives that are currently employed should be worried due to short sighted budget cuts thinking that AI can replace them, but these industries will quickly learn that this "AI" has no real longevity and will continue to produce worse results as time passes. This "Generative AI" is inherintly non-generative. This AI amounts to essentially having every pixle or letter (depending on what it's being used for) weighted against millions of pieces of data and the algorithm very quickly allocates what it considers to be the highest valued pixel/letter based on the input data. The models can't get much better than they already are without WAY more training data, this method of training is becoming more and more difficult, and it takes real artists a long time to actually create something new. The point being, "AI" is about as good as it can get right now, and that's not very good. It can convince some people that it's the way of the future, but most people know it's a grift and that's why so many companies are jumping on board. For a while, just having "AI" as something your company worked on/with was enough to boost your market value. Now, the inverse is true and people recognize that "AI" generally means something is of inferior quality. I don't think we're ever going to fully get rid of AI because it has tricked too many people, but the chances that anyone realistically attempts to use AI for real work in the future is pretty negligable. I just hope that the artists of today don't suffer too much in the mean time
Just reading the comments people are so negative on this as if it is a finished product. Your comment is the first one that actually had some logical sense that I’ve read. This thing is just a baby step towards a proof of concept. It’s meant to open up all the issues so it can continue to improve.
When I saw the initial Stable Diffusion and GPT models, I made the classic blunder of forgetting the slippery slope, seeing it at the time as an ephemeral novelty. Never mind art or employment, now I'm starting to worry we're entering into an incoherent hyperreality as we drown in the ocean of AI garbage that is soon going to comprise the world we perceive.
hot take... The people who over invest in ai in the wrong ways will put themselves out of business. Gaming will be fine, existing public companies in the space will die tho. Indie devs using local gen ai to aid workflows will catch up to triple A in the next few years.
Nothing made by generative AI has surprised me and evoked in me such a visceral feeling as this AI model does since those old "name one thing in this photo" images from 2019. For me, they're some of the rare examples of something that perhaps only generative AI technology could have made, and almost everything I've seen since then has been more insipid and forgettable, becoming more convincing only as it becomes less interesting. I tested out the model myself after watching the video and enjoyed looking at the sky or the ground for a moment and then looking back, not knowing what I'd see. It genuinely thrilled me when - just for a few seconds - it would fall into some unstable equilibrium which I would try and fail to hold onto, like forever falling from the sky into the ocean, or teetering between two wholly different interpretations of the same blurry image. Even just walking around and seeing how it will change its mind on the basic geometry. Perhaps my visual cortex has been exposed to so much Minecraft that it messes with my sense of reality, while avoiding the most disturbing artifacts of AI trying to depict photorealistic humans. Trying to play it like actual Minecraft as if there is anything behind the immediate image being shown is bound to be frustrating. But that's a problem for the developers trying to sell it as enabling future "consumer products", not for me. I wouldn't want interactive AI video generation models to replace video games any more than you do, but as you say at the end, I'm so skeptical of that ever happening that I can't help but feel you're a little harsh on the "game" itself, when it may turn out to be one of the only interesting things to come of this. There are other ways generative AI might impact game development (and already has) which I'm more concerned about, like use of generative AI to come up with mechanics, narrative ideas, dialogue, assets etc. The problem with AI minecraft imo would lie more with the end goal, the hype, and the motives of companies pushing it. I don't think that should stop us appreciating Minecraft AI for whatever it is, whether or not it does anything for you personally.
The idea of tricking the game into sending you to another biome by looking at blocks from that biome is an awesome idea. It just can't be the whole game.
So, this tech has absolutely no future. It's catastrophic. BUT I found it oddly fascinating how close it is to how your brain does things when you dream.
Looking at this nightmare is such a testament to the limits of AI models because there are- *THOUSANDS* upon thousands of hours of Minecraft footage they could have trained this on, and I'm sure they did, and this horrifying noise soup is the result
@GreenBlueWalkthrough Photography, CGI, and Painting are all art but fall under different mediums. Each exists without the other. Gen AI can only exist with a database, and 99% of the time is trained off of stolen/uncredited work. People will argue that the techs whole purpose is to make art more "accessible," but paper and pencils are much more accessible than an expensive monthly subscription to a service.
Don't worry, it will eventually destroy every other job and hobby too! Welcome to the post-truth future we are burning our current society and culture down for, all because a handful of people who already have basically infinite money want to have infinite +1 dollars.
@@andyyo9554 AI will to many things, but it won't lead us to a post truth future. If anything, it'd motivate more people to get up, go outside, and investigate themselves when they hear about something. I'm seeing normal people right now going "don't believe that, its probably AI", so I have no reason to believe they'll suddenly 180 and uncritically believe the AI.
@@luckyducky7819 honestly I think it’ll just lead to people dismissing everything they disagree with as ai and saying everything they do agree with is real Obviously reality intrudes to a certain extent but for the most part I think this will just make it easier to silo people into different info bubbles and pit them against each other There were already so many people buying into networks of obvious disinformation and nonsense before ai, and most of them seem to be willing to believe anything if it’s said with conviction and confirms their pre-existing beliefs My deepest fear is that it’ll become easier to cover up actual atrocities by saying the evidence is ai, especially in places with heavy government information control (and it seems like a lot of places are headed that direction…)
This reminds me of actual dreams. Like, you’re in school walking to class but somehow the stairs completely skip the second floor and you have to walk down the third floor hallway and leap over a railing and climb down to the second floor and then you have to shimmery through a miniature version of the hall and slip into an open field where people are having lunch, and then you realize you’re in your underwear but it’s okay because you have an invisibility cloak. You just have to be careful and not be noticed by the pterodactyl.
I haven’t finished the video yet but if I could describe playing this game it would be like playing a dream (or nightmare) of Minecraft. The images are hazy and you don’t really have control over anything. And when you wake up you don’t really know what just hit you.
I can’t finish the video. Not because the video’s bad. But because the game is so bad it makes my blood boil. Who releases this kinda stuff? People with really thick skin.
This is some of your best writing to date. What’s incredibly sad to me is that AI should be used to fulfill menial labor and free us to pursue more creative endeavors. Who knows what kinds of art could be produced with that free time.
Though for that to happen people would need to have money for their basic needs. That can’t work when most American families are living paycheck to paycheck
@ Totally agree. The use of AI for labor purposes would require our incentive structures to be focused on people rather than profit. We don’t live in that world, but I do believe it’s possible.
The first minute & 4 seconds alone is already the best explanation for what “AI” is actually doing I’ve seen so far for real. The ease at which autocorrect is inadvertently trained to “correct” my correctly spelled words into my usual typos immediately comes to mind, applying such homogenization everywhere else is just… it’s indeed very different from how the original Minecraft or even the one flight-simulator’s graphics were programmed. Humans made computer generative technologies better, how hard billionaires are trying to simply take the humans out of the equation is very telling. I’m normally not a fan of sitting down and listening to entire rants on this subject, but this is one to watch from beginning to end. 👍 It goes in some super neat & unexpected directions, kudos.
when you transition from using the oasis footage to showing actual minecraft it looks like those "rtx on/off" comparisons with how much more sense the visuals make
Everyone trying to shove ai everywhere feels like the sort of thing where... the people with all the money bought into the hype early on and now that they're seeing how unprofitable it is. Are trying to shove it wherever they can to try and get their money's worth.
As a game developer, I don't feel threatened by AI. Not because I think it will "empower developers" or any ridiculous nonsense like that, but rather that all the AI slop we've seen so far just doesn't impress me. Oasis is a quaint little tech demo to show to investors, but by many definitions it doesn't even qualify as a game, let alone a GOOD game. Evangelists will say "but the technology will improve over time", but there are some constraints baked into this method of AI generation that makes creating a game literally impossible, a primary one being persistence. Game worlds need to be persistent, especially in a game featuring building structures as a primary mechanic, and AI models are incapable of large scale persistence. The more work you do to try and introduce persistence, e.g. adding models and textures that the game generates and saves to be loaded later, the more it starts to resemble a traditional game engine. Eventually you will get to a point where it's just worse procedural generation. There's obviously a whole host of other problems-like how there are no actual mechanics in the game, only hallucinations of mechanics-but if I went into detail on all of the failings of AI in gaming, I would be here writing this comment for eternity.
I personally have nothing but vitriolic contempt for this style of large learning model ai for a number of reasons, but if there is ANY tiny sliver of a silver lining here it's that this garbled mess is probably the closest you could visually represent dream logic (as constrained by minecraft but the metaphor still holds.) If I were to describe a dream sequence in say a TTRPG to a player or write it in a piece of fiction, the level of shifting, melding, blurry, almost but not quite recognizable events happening to the subject would fit quite nicely.
This is the equivalent of using the first computer ever made and shitting on it's ability to be improved in a useful or desirable way. Look at AI development, it doesn't take too much thinking to realize that this is the worst it's going to be at generating games. Your money talks and companies will compete for it. Ai slop games will be a thing for sure but so will actual games built with AI in a creative way. AI doesn't have to be the source of creativity for the things that come out of it. What AI can be is the tool creatives need to reach their visions with ease allowing them to have complete control over what they are trying to bring to life.
because the way first computers worked is that they gave probability-based output, and the way we improved them is by making sure they could calculate probabilities better? big difference between "core idea is good, but the hardware is massive and inefficient" and trying to make the AI guess better. i mean on some level you're right, considering how much stuff gets called "AI". but i doubt that it will be an iteration of the current technology but just improved, it will probably work completely differently from current "AI"s.
It's such a weird time. My experience with it had me trying to manipulate the AI. Once I knew what it was doing, I could wrestle with it and 'build' a house by looking around in certain ways. If this video learning can be combined with some kind of mechanics learning, we'll be seeing some WILD stuff. There's going to be an era of 'games' that are just LSD Dream Emulator turned to 100x.
geez from the looks of it, this is what old folks used to say that "videogames are just a waste of your time", i mean looking a this example i totally get what they meant by that, this whole AI games craze only creates what look like time wasters, instead of a interactive entertainment medium,
As a game dev myself, I struggle to see any potential utility in AI, even if it were "used for good" or whatever. The basic fundamentals of how it works makes it either useless or reliant on theft. Like what artist is going to bother uploading their entire body of work as training data just to get an AI model that shits out half-measures with absolutely 0 new life? Having no control at all over what it even gives you. What voice actor is going to be happy replacing themselves with the garbled phantoms of their prior performances? The tech as it exists just simply isn't useful to artists past maybe a smart fill tool in photoshop, or generating some surreal audio samples for experimental music...
This is exactly how a nightmare feels. It starts out in recognizable territory, things start shifting and morphing without your control, and it degenerates into nonsense and noise as you frantically try to assert some form of control.
i think its great that people focus on AI as a labor exploitation issue (it is) but it can sometimes distract from the fact that AI is also inherently worse at most jobs than humans
I think if AI videogames were to take off, it would more be a case of the developer only needing to create graphics to a very low, simplistic level, but then overlaying an AI model that takes in what the game engine renders, and creates something far more visually realistic, while not having to worry about problems such as object permenance, or non-euclidian geometry that comes with fully ai generated experiences.
While the idea of a playable nightmare is somewhat fascinating to me and some advanced level of procedural generation would be necessary for that, yes to everything said here. Best wishes to those affected before things start getting litigated and, hopefully, better.
"We made a shittier version of the most popular game on earth and only needed milions of hours of footage" is not the win they think it is. The tech will get better, and the useful parts will probably be adopted by people once the copyright/plaigiarism issue is sorted out, but the main issue with AI still stands, it cannot create something new, it can only make derivative variations of already existing stuff. I hope the AI bubble crashes soon so the hype can die and the money and man power can go to something actually useful, but i think we will first see mass adoption, before a mass rejection
The discord contest on the side of the screen is so disingenuous that it's actually absurd. Genuinely after 5 minutes of experimenting and testing the limits of the "game" I'd be shocked if there was anybody gullible enough to think that any of those "challenges" were actually possible. "1. Kill a zombie and have rotten meat drop" There are no zombies. There is no day night cycle and there are no caves so there is never a point where a zombie would spawn. Even if you did somehow get the "game" to generate a rough approximation of a zombie actually killing it would be a nearly impossible task as the game has no concept of hit points and getting it to drop any item, let alone the item it's supposed to drop, is simply not on the table. "2. Explode a creeper" Same problem as 1. There are no creepers. "3. Get shot by a skeleton" Not only are there not skeletons but even if there were I think the sun would genuinely die before this thing would ever allow an arrow to fly through the air and hit anything without turning into a distant block of dirt first. Not to mention actually getting any mob to even recognize a player exists is just not going to happen. "4. Punch the ender dragon" This one is just laughable. There is no ender dragon. There is no end. There is no end portal. There is no strong hold to contain an end portal. There are no caves to contain a stronghold. There are no eyes of ender. There are no endermen. There are no blazes. There is no nether. There is no way you could ever even take the first step towards accomplishing this task and implying that it would ever be possible for anybody at any point is just laughable. At this point they might as well just add "become the first man to walk on pluto" to this list because it is genuinely more plausible that someone would do that than it is that they would ever even see an ender dragon in the current state of this "game".
With nearly everything the tech is so so so cool, but all the CEOs and financebros want to turn it into money and just sacrifice a whole workforce which is very sad
I see some potential in this Technology for a Cosmic Horror Game, if one could balance AI Generatet Scenes like these against a solid core Gameplay, something Magical could be created.
Machine learning (which is all that this "AI" stuff is) is extremely limited in what it can reasonably do. With time this will become more evident. Simple math problems like adding two numbers together requires an insane amount of training to become somewhat reliable. With a traditional algorithm it's one line of code. So is it really better to train a model to learn how to better guess an answer or just have something you can rely on 100% of the time? In the realm of games, video game systems work exactly because they are deterministic and follow strict logic. You learn from the game, not the other way around. Probabilistic algorithms have their place in games like, as you put, procedural generation and such. But the core systems must always work off of deterministic logic if you want to make a fun video game, and there's a reason "biomes" are a thing Minecraft to make the randomness more interesting. Getting these probabilistic models to resemble deterministic systems is such a huge waste of resources and time. They can, by definition, never be as good as the real thing (something made by a human).
The discussion in this video was fascinating and galvanizing, but watching that "gameplay" was outright repulsive and strangely and deeply uncomfortable. Hamish, I feel like you really took one for the team putting in the time playing that thing and putting together this video
So when you press w, it essentially sends a "move forward" prompt to the AI video model and shows you that. Lol. Maybe it could be some kind of artistic novelty one day. It's a shame the economic conditions in which we live with ai make this hard to imagine though.
This looks bad enough that it might actually turn people off of AI in games? I always feel bad for Tommy Thompson who studies AI on gaming but how it is/can be used in a fundamentally different and beneficial way than how it's a being used by corpos currently. I always feel like I'm inadvertently shitting on his work when I go off on AI in games such as this one. Being a proponent of something that gets the worst part of it co-opted by capitalism has got to feel awful.
Makes me wonder if it's calculated investor bait. If the money folks aren't knowledgeable enough to judge it as a game, they may just look and see what traffic it's doing on social media. And hilariously unplayable can generate a fair bit of coverage and memes. At least for a while.
I really cannot stand the ways we have been "utilizing" these "advancements" in all of the worst most useless ways... Rather than finding ways they could utilize it to improve or benefit different things.. They have just forced in this trash to turn everything into useless slop. Even previous systems that used to function well have been ruined by this "advancement of AI nonsense"
Because it’s all about the “potential” from that context, it doesn’t really matter that it looks like shit-it has potential. This is how a lot of these companies are able to sell people on this shit: buy into it now and it will eventually bet better.
The very nature of the tech, Let Alone the need for reference and what that means for copyright and intellectual property laws, means this entire thing is just untenable and the tech Can Not do what they're promising it can do. And Yet, it really be like that tv show by Mike Judge, Silicone Valley, where companies are racing against each other to create and bring a fabled tech to market. And apparently the real Silicone Valley does this kinda thing all the time.
honestly? ai generated games wont be engineless. i Seriously doubt that. at most? still dependant on an engine, but able to generate each component live if necessary. the most benefit i can possibly think ai in games can give is better npc's and making those companies actually make dynamic worlds again
I wouldn't worry about gaming's future. They'll try to sell it, everyone with half a brain will recognize it for the trash it is, and it will crash and burn in the end as people increasingly realize that this "AI," is about as far from what sci-fi has promised (even on the low end of the AI authors have dreamed up) as an abacus is from the internet.
My prediction is that AI will stay (sorry) BUT that'll be seen as the low class version of production, like how McDonalds, though hugely profitable, is generally seen as unhealthy, cheap (yet still overpriced), and tasteless. This doesn't mean games will become "artisan", but I can see games, especially indie games, being marketed as of higher quality and "hand-made", as it were. Yes, AI games will just need to sell a few copies to turn a profit since expenses will be so low, but there's a reason why McDonald's is the butt of the joke, and why they'll never break through that perception, no matter how much they renovate their design or try to appeal to more mature tastes.
Here’s a cynical take: it honestly doesn’t even matter if AI can get good enough because that’s not its purpose. Its purpose is to be “good enough” to justify getting rid of talent (not that the industry needs any more of a reason to let people go) and allow corpos to shit out product faster and faster. But even worse: I think there’s a large contingent of people who just don’t give a shit. Don’t care about art, don’t care about the people who make it. They just want the thing and it being somewhat decent. And that’s enough.
Yes, it looks awful. But you don't seem to understand the step that this represents. This is like the first ape sparking fire with two stones. Yeah, primitive, rudimentary, lacks style, won't be able to repeat it twice. But we always underestimate the rate of revolution when we are in the first T+1 moments.
The scary bit is that this, while it sure does look awful, is the worst AI will ever be. Everyday it gets incrementally better and that future we all dread gets a little closer while we are busy laughing at the initial, shitty results.
Or maybe the technology fundamentally flawed and the promise of it eventually getting ‘better’ is just to scare, or ‘disrupt’ stable industries into adopting new tech in the hopes of cornering a new market.
@@GreenBlueWalkthroughthere is no AI art as good as human art. To suggest otherwise completely misses the point of art. Art is human expression and creativity, something a machine is not capable of. The AI art you love so much IS human art, because it’s stolen and copied from the work of human artists. AI cannot create something original. It would be like me tracing a drawing and proclaiming it as my own, only worse. A five year old’s crayon drawing of their family where mommy and daddy are bigger than the house is a better piece of art than literally anything AI has ever created.
As the AIs fill the internet with their slop their own training data will be corrupted, they could infact get much worse, and we only need to wait until this bubble bursts as tech companies building whole power stations for this stuff realise it doesn't actually have a means of making money
@@eamonntee Look at how dumb and silly the first AI generated images looked vs the practically perfect output they have these days, to the point that it has us worrying about the implications of its use for disinformation. In short, it has already happened. But maybe you're right, that's as good as it will ever get for AI and they're using that as proof for more funding.
I kind of think your outlook on this is kind of extreme and lacking in nuance. I personally think that AI isn't some big bad thing that's inheritly evil, it all depends on how it's used, I feel like it has massive potential to advance humanity in all kinds of fields, the fact that it understands specific context is also very useful for specific questions in general and it can be used as an awesome tool to learn and even aid creatives, for example, a writer might use ai to correct spelling and punctuation errors, a small budget film production crew might use ai to generate crowd shots that would otherwise completely break the budget, in general it allows smaller teams with tight constraints to put out increasingly larger things, we're steadily lowering the barrier between the concept and the finalized product, hell at some point with the inclusion of brain implants we might be able to completely cut out the middleman and just export stuff directly from the imagination, this of course opens up a whole other can of ethical concerns. Of course it can take away jobs, but eventually this will impact everybody and we might need to come up with new societal systems now that the world basically runs itself, this doesn't just apply to creative areas either as the amount of fields that it will be able to automate is steadily increasing, at the end of the day people are just going to do stuff because they enjoy it, not because they have to do it for the infrastructure to continue functioning, and for creative fields there will surely always be people that want to consume human work. I saw a great comment on UA-cam at some point in regards to the progress of image generators, I think it went something like this: "In retrospect, the fact that it looked like anything at all should've been a sign of what was yet to come."
I think you have a very idealized, perhaps even naive view of what is actually happening with AI on a practical level. The problem comes when you actually look at who is pushing AI the hardest and what they stand to gain by it becoming widespread. Right now, Large Language Models do have specific use cases where they not only make sense but actually help people get specific jobs done. The problem is that billionaires tech industry types see it in the same way they saw NFT's and even social media platforms. They just want to "disrupt" established arts and industries for however long they stand to make a profit by doing so. It isn't about some kind of technological revolution, "democratizing art" or some other simple-minded catch phrase. It is just about extracting profit for as long as possible and then jumping to the next irresponsibly developed technology to exploit. There are some specific things in your comment that I want to unpack a bit. The first is the idea that LLM's can "lower the barrier between concept and finalized product". This is true in the sense that AI can be used to generate large scale content without any major work or resources on the part of the prompter. That being said. It isn't going to suddenly empower the amateur to create better art, it will just enable them to create larger scale content. What's more. It will create a sense of false capability where amateur creatives don't learn important skills that can not only make their content better but also enable them to better utilize LLM's in a way that truly makes the process better. Let me use music as an example since the video brought it up. I have seen some of the more devoted AI enthusiasts talk a lot about how AI generated music will "democratize the process for everyone" and will suddenly let people skip the "boring" and "overly difficult" process of learning how to play an instrument, sing, or even use existing technology (like drum machines and the like) and just get straight to "being a musician". The problem with this is that music was already democratized. It has been democratized since someone could buy an acoustic guitar for a relatively low price back in the 1800's and early 1900's. It happened when folks started singing songs as a family or as a community. It happened when getting recording time at a studio to make a demo could be had at attainable prices. Right now. A teenager working a entry level job has the option to buy inexpensive audio interfaces, mixers, microphones, and software that will enable them to create very polished music. That same teenager can also buy very good instruments (guitars, keyboards, etc) for low prices and utilize free resources to learn how to play them. The notion of "democratizing music production" has already happened. it is a problem that has already been effectively solved. The only thing LLM's can add to that process is generic sounding content created via prompts. It won't help the amateur prospective musician become a musician but it will help big companies like Spotify push cheap, LLM focused content on their platform since they won't need to pay artists for it. Lastly. Let's talk about the "taking away jobs" thing. It will happen and the fact that there is such a malicious hunger for it in the AI enthusiast circles is alarming and worrying. That said. There will not be a viable replacement for those jobs. People will just get put out of work and there won't be any viable alternative. This is what ALWAYS happens historically. We never have that magical thing where a job is automated and some other equally viable job magically pops into existence for displaced labor to move to.
@@startrekmike Great points, i considered mentioning the billionaire thing in my comment but thought it would kind of make my comment more conveluded. Yeah of course its primarily billionares that are pushing this the most, however it's not like it's just billionares, at this point there are already a bunch of AI's that are open source and made by smaller teams of people, pandoras box has been opened not just to the giant cooperations, but also to the people, sure the most cutting edge systems will most likely always be closed source but the open source community and smaller independent teams are making good progress, ai can be developed in all sorts of directions by different organizations that range widely in size. Regarding your points about the barrier, i think you're right, it could definitely lead to amateur's to not learn certain essentials, what i mean is that in the future budget or scope in regards to finializing a project will probably become less and less of a concern, which i think can definitely be a good thing, i do think that people who are passionate about something will think about what kind of knowledge and skills they are aquiring through their usage of AI and acommodate. About your music example, i agree with you that music is already democratized, i think the main idea is lowering the gap between idea and product as mentioned before, the problem with this is of course the certain lack of control that you currently have over the outputs of AI like this, i still think that it does have a place to exist though, for example it allows people that aren't in the field of music to quickly get original music that is tailored to the specific scenario that they need it for without having to take a detour to learn it or hire an artist even though they might not even need it that often. It could also be used for potential real-time generated music that dynamically reacts to a players actions in a game for example without ever repeating. But yes, what you mentioned is definitely already happening, i feel like it all has its potential ups and downs. I do realize that the perspective that i voiced here was pretty optimistic, i acknowledge the fact that in the wrong hands this could go very, very wrong, exctinction levels of wrong, but i think that that is very unlikely to happen and also that it has the potential to do a lot of good, even if the reason certain individuals are pushing it might be self-centered. At this point the cat is out of the box and we just have to deal with it and make sure we're not digging our own graves, i believe that this can end well, there are going to be problems but i think this will not be catastrophic and humanity will prevail at the end. And lastly, i agree about the displaced labor, but this is on a scale we've never seen before, combined with consistently improving robots almost every single job could be replaced at some point, and it will only get better, eventually it might even just improve itself automatically. This will most likely cause so much job loss that eventually traditional jobs might not exist anymore at all, like i said we will need new societal (and economic) systems in place to account for this tremendous shift, we might just be making stuff up at some point to keep ourselves busy and the economy functioning or in a perfect world people might just follow whatever kind of indeavors they want without any reasoning behind it other than doing it for fun and we might need to rethink the entire economic system.
@@toyo8460 When I think about the displaced labor issue, I can't help but think about what happened when we shipped a lot of our manufacturing overseas or just automated it. These jobs were genuinely good and often gave High School graduates an opportunity to get a stable career and start meaningfully saving pretty much right out of the gate. These were the kind of jobs that created that post WWII image of prosperity. When those manufacturing jobs withered away, the ones cutting those jobs to line their own pockets were also selling the American public on the idea that something just as good would fill the void. That instead of manufacturing jobs, there would be tech related jobs and administration related jobs. It never really happened that way. Instead of the good manufacturing job market getting filled by something equal or better, it was instead filled by retail and service based minimum wage jobs because that was really the only other large scale employment opportunity available to anyone who didn't have a (largely unnecessary) college education. This brings us to the next tragedy. We never really adjusted culturally or systemically. We just kinda went on assuming that there were these great bread and butter career jobs for everyone and pretended that those who were now stuck in retail/service deserved to be stuck there because they didn't pull on their imaginary bootstraps hard enough. We still haven't really acknowledged that we don't have enough good jobs and now that we are getting ready to give AI a bunch of what is left, do you really think we will change our pattern? In regards to the idea of using AI for art related jobs. One thing you really need to keep in mind is that AI isn't going to really do much of anything to established, famous artists, actors, or musicians. Instead, big companies are going to use LLM systems to generate the low-level, day to day stuff that we as consumers don't really think about. Here is the problem. That entry level stuff is how someone goes from being a hobbyist to a paid artist. Those kind of jobs are how people build themselves up enough to actually become those famous, well-known artists, actors, and musicians. AI is going to close that door and there will be no opening it again once it is shut. We are cutting off our nose to spite our face in that regard and the only winners are going to be the CEOs who stand to make a lot of money in the process. Lastly. We are already in the "worst case scenario". Billionaire tech CEOs are already the forefront of the AI push and they have armies of influencers on social media who are trying to make sure that everyone else is enthusiastic about getting screwed over. It is scary how cult-like it has become but after seeing the same thing happen with cryptocurrency, I can't say I am overly surprised. LLM based systems were fine when it was just something that was being used in very specialized roles and largely out of view of the tech CEO's. Now they see it as their next big hustle and we are all going to pay.
One thing I could see AI being really good for, is generating side characters in RPG’s. Imagine if you could use AI to give every single “extra” character a personality or backstory. You’d never have NPC’s named “villager” that do nothing besides spout one or two lines and walk around raking fields all day. They could have genuine friendships, rivals, love interests, radiant quests etc. and everything about them could be generated by a well programmed AI. It’s way too much work for an actual dev team to do this which is why it never happens. I think that would actually add a lot to a game, whereas most ideas regarding AI seem to be about outsourcing work and passion to a passionless machine. It should be used to fill in gaps that were going to be gaps anyway, not to create gaps to be filled in by something worse.
Oh my god 80% of this video is an unhinged rant at what is basically just the equivalent of GPT2, what are we doing here, this is embarrassing from an essay point of view, the only real take were the 2 minutes at the end. I'm not even against the point, it's just so badly made. Are there really people who pay his Patreon for this lol
I initially misread the thumbnail as "ALL GAMING IS AWFUL" and thought it was a pretty bold turn for you.
The most searing take of all
I'm a 19-year veteran of the games industry (im 40 now), in AAA and indie - worked at Ubisoft, Rare, EA, you name it! The only people clamouring for genAI are, sadly, the top floor executives with zero actual development experience. They see Magic Robot That Creates Anything You Desire and go "ka-ching!," without considering for one second how the model was trained, whose work was stolen to train it, and quite importantly how long this wild west can continue before governments start to litigate on intellectual property ownership rights extensions, compensation for targets of training, and more. Effectively, whilst the technology will improve, that's not to say every country will permit this unimpeded data grab/IP skimming forever.
Ultimately I see it like this, bud - Pot Noodle and Spam did not kill off restaurant dining. GenAI fills the niche of the "machine-made slurry of questionable origin which doesn't overly impress nor satisfy, but does the job in a pinch." Like spam, some will see that as good enough to consume every day with no regard beyond a full belly. Others will see it as a last resort - a "when all else fails" option. And others, and I'd argue a fairly decent number, they'd simply rather spend a little extra now and again to sit in a person's restaurant with a loved one, and dine knowing it was hand-crafted with love; to enjoy each carefully-crafted morsel and refer back to them fondly when they do have to eat a Pot Noodle a week before payday.
I don't see GenAI as replacing me for the games that I make nor the customers I make my games for. I see GenAI as making for a customer who only sees content, not craft. Who doesn't care if the eggs are free range or not... they just want the cheapest eggs. Those who don't care if Nestle steals water from the poor and needy, because they want a kit kat.
Simply put, those who respect beyond the surface level will always want human-made games. We may lose some customers overall... but those were just people who found cheaper Pot noodles and spam.
19 year old veteran? as in having worked in the industry for 19 years?
@cataclystp haha, yes - in the industry for 19 years, I'm 40. Man, I WISH I was still 19 though
You're playing a dementia patient's memory of minecraft.
Every block everywhere at the end of infinity
@jackiespaceman Exactly. Except that album is fire and this is soulless garbage.
Exactly this. I said it was like a dream. But dementia works too lol
Weirdly enough there is an actual Doom mod with that concept. It's surreal horror at its finest.
@bencesarvari2235 You say that like it's a bad thing, but I think that sounds like a super interesting experience. You sold me on this. im going to try this out.
The longer the AI hype cycle continues, the less I'm worried about it when it comes to art, honestly. The comparison to NFTs, I think, is very apt. Something that blows up, grifters looking to make a quick buck (or a lot of bucks) dive all the way into it, and then people catch on that its got no clothes and the market for it crashes to nothing. People talk about how "amazing" AI is, and how good it's gotten recently, but I've never seen a bit of it that didn't suck. "But it can make videos that..." yeah, I saw the woman walking down the street. I saw her jacket randomly morph when some of it was outside the screen for a moment.
The fundamental problem with AI, as it stands now, is that AI doesn't know anything. It doesn't know that 2 + 2 = 4. If you ask it for the answer to 2 + 2, it calculates that many other people have said 4 in the past, so it says 4. But it doesn't know that. This may seem like a pedantic distinction, but it's the core of why AI is so bad outside of tiny little windows of use, and why even in those tiny windows it needs to be watched like a hawk so it doesn't do something like tell you to Elmer's your pepperoni to your pizza. I don't think anything has demonstrated this so clearly as AI Minecraft. Just think about what had to happen for MAInecraft to exist:
1. Create Minecraft
2. Minecraft gets insanely popular
3. Minecraft becomes one of the most-streamed games of all time (or is it THE most-streamed game?)
4. Tech company spends probably millions of dollars to train a model on Minecraft videos
5. Tech company releases a broken mess that can't remember anything that happens offscreen because AI doesn't know anything
"But it's going to get better!" people like to say. But, let me tell you a secret: it won't. Literally, the LLMs are out of data to hoover up; they're as good as they're going to get. And they're not good. But more importantly, the fact that AI doesn't know anything is the structure of the system. It's not something that can be "fixed". Someone would have to invent an entirely new technology, at which point we're talking about something else.
But also, look at what it took to make MAInecraft. Look at step one! The only way to make a "game" with AI... is to have already made that game! Not just already made that game, but it has to be a simple game, with a very consistent UI and look, and millions of hours of available video to suck into the model. And then, the AI can spit out a Cronenbergian nightmare that, in some single frames, looks similar to the original. Yes, they'll reach a point where they can say, "Make Minecraft, except replace the sheep with facehuggers from the Alien game I also trained you on". But that just means they have to make TWO games before they can make their one bad AI game. And the more complex the game, and the more you feed into the model, the less the AI will be able to handle it. Just look at the inventory in MAInecraft. Someday, I'm sure they'll get that into the model in a usable-ish manner (except it won't remember what items you have because it can't know anything). But imagine if the model was trained on a hundred different games, and you tried to "prompt" the UI into a usable system, and "prompt" how inventory worked. The AI can't handle that. And the poor employees that are ordered to make it work will try, I'm sure, and get fired when they can't deliver, I'm sure. But, quite simply, the technology is not able to do it.
There are plenty of areas that I'm worried about AI. Surveillance, law enforcement, internet search, driving, propaganda, etc.. Mostly, because it's going to be "lying at scale" and people are going to rely on those lies and die (see: mushroom foraging as written by AI), and/or because it's going to remove accountability from a critical decision (no person crashed the car, the AI crashed the car). Art will be as fine as it ever was. Artists might have a hard time while foolish bosses fire them expecting AI to fill the gaps, but AI will not fill the gaps. Nobody's going to buy MAInecraft, or the movie version of it, or the novel version of it. Because it's bad. Even images, which is arguably where "AI art" is strongest, the result is very samey, very soulless, very obvious. The people wanting it didn't want art in the first place.
To take Adam Savage's words and treat them like my own: Art is about presenting a point of view. A machine does not have a point of view, so what it makes is not art.
📌💯
Extremely good points. Thank you.
On your point about creating an entirely new technology, they would still call it AI. This isn't AI either; they call this language model stuff AI for marketing purposes. So if they created an intelligent AI, it would still be called AI, my brother.
@@aldiascholarofthefirstsin1051 They would call it AI. That changes literally nothing about what I said. The point is that the technology to do what people claim LLMs can be "improved" to is not possible with the current technology. Something new, which does not currently exist in any form, will need to be invented. The names of these things is irrelevant; what's important is the technology.
It'll burst, and then it'll surge back.
They don't even have the common decency to call it "M-AI-ncraft".
Hearing this in a Rich Evans voice:
Maaaaaiiiiiiiiiiiiiinecraft!
Seen some AI minecraft. Person was climbing a hill and once the hill filled the screen, the AI's lack of object permanence kicked in and decided he was looking directly at the floor.
E: Fill screen with sheep. It is now a mountain side.
The developers of Oasis had to forego most of the context buffer to make the experimental project run more smoothly. This thing is really more of a novelty/proof of concept than a game.
The most impressive thing about AI minecraft is how it managed to perfectly encapsulate what it feels like to navigate a Nightmare.
a Dream*, a weird one. A nightmare would be worse.
this is what it fells to sleep on antipsychotics
This is how Alan Wake has been playing Minecraft in the Dark Place.
It's really impressive that a computer can (sort of) simulate Minecraft without actually playing Minecraft, but what's even the practical application? Nobody wants to play the weird incoherent LSD Dream Emulator version of Minecraft, it probably takes way more tech to actually run than the real thing, it's nauseating to look at, and completely unplayable as a game. So. Who is it for? What purpose does it serve? Why were so many resources put into making it? Who will even care about it? It's like spending billions of dollars and hundreds of thousands of hours of research and labor to develop a 5 hour long process for putting on your shoes.
Honestly, there is a great artsy indie experience with dream logic to build with this concept.
if wasn't soulless company garbage
What I hate about Silicon Valley's promotion of generative AI, is that instead of seeing what it does well with and applying it to something useful and social good, it's just being shoved everywhere hoping to be the first to profit. Billions have been invested and there hasn't been much ROI, they're getting more desperate as time goes on.
A great use would be generating NPC banter as you walk through a city. That way it wouldn't repeat and it's not critical path content.
As someone who works as a programmer and did some work with AI in school, I can safely say that I'm not really concerned with AI in the long term. Artists and creatives that are currently employed should be worried due to short sighted budget cuts thinking that AI can replace them, but these industries will quickly learn that this "AI" has no real longevity and will continue to produce worse results as time passes. This "Generative AI" is inherintly non-generative. This AI amounts to essentially having every pixle or letter (depending on what it's being used for) weighted against millions of pieces of data and the algorithm very quickly allocates what it considers to be the highest valued pixel/letter based on the input data. The models can't get much better than they already are without WAY more training data, this method of training is becoming more and more difficult, and it takes real artists a long time to actually create something new. The point being, "AI" is about as good as it can get right now, and that's not very good. It can convince some people that it's the way of the future, but most people know it's a grift and that's why so many companies are jumping on board. For a while, just having "AI" as something your company worked on/with was enough to boost your market value. Now, the inverse is true and people recognize that "AI" generally means something is of inferior quality.
I don't think we're ever going to fully get rid of AI because it has tricked too many people, but the chances that anyone realistically attempts to use AI for real work in the future is pretty negligable. I just hope that the artists of today don't suffer too much in the mean time
That is a tech-demo. I don't think it was supposed to be better than the real thing right now.
Just reading the comments people are so negative on this as if it is a finished product. Your comment is the first one that actually had some logical sense that I’ve read. This thing is just a baby step towards a proof of concept. It’s meant to open up all the issues so it can continue to improve.
"There is only so far that company can realistic push this"
Are you sure there is a realistic limit they can push this?
When I saw the initial Stable Diffusion and GPT models, I made the classic blunder of forgetting the slippery slope, seeing it at the time as an ephemeral novelty. Never mind art or employment, now I'm starting to worry we're entering into an incoherent hyperreality as we drown in the ocean of AI garbage that is soon going to comprise the world we perceive.
I'm a game dev and I enjoyed this project. I get it's shit as a game, but it is definitely impressive.
Cool? Sure. Impressive? Not even slightly.
hot take... The people who over invest in ai in the wrong ways will put themselves out of business. Gaming will be fine, existing public companies in the space will die tho. Indie devs using local gen ai to aid workflows will catch up to triple A in the next few years.
I love your outlook on this. Protect workers, protect people, protect humanity.
Nothing made by generative AI has surprised me and evoked in me such a visceral feeling as this AI model does since those old "name one thing in this photo" images from 2019. For me, they're some of the rare examples of something that perhaps only generative AI technology could have made, and almost everything I've seen since then has been more insipid and forgettable, becoming more convincing only as it becomes less interesting.
I tested out the model myself after watching the video and enjoyed looking at the sky or the ground for a moment and then looking back, not knowing what I'd see. It genuinely thrilled me when - just for a few seconds - it would fall into some unstable equilibrium which I would try and fail to hold onto, like forever falling from the sky into the ocean, or teetering between two wholly different interpretations of the same blurry image. Even just walking around and seeing how it will change its mind on the basic geometry. Perhaps my visual cortex has been exposed to so much Minecraft that it messes with my sense of reality, while avoiding the most disturbing artifacts of AI trying to depict photorealistic humans. Trying to play it like actual Minecraft as if there is anything behind the immediate image being shown is bound to be frustrating. But that's a problem for the developers trying to sell it as enabling future "consumer products", not for me.
I wouldn't want interactive AI video generation models to replace video games any more than you do, but as you say at the end, I'm so skeptical of that ever happening that I can't help but feel you're a little harsh on the "game" itself, when it may turn out to be one of the only interesting things to come of this. There are other ways generative AI might impact game development (and already has) which I'm more concerned about, like use of generative AI to come up with mechanics, narrative ideas, dialogue, assets etc. The problem with AI minecraft imo would lie more with the end goal, the hype, and the motives of companies pushing it. I don't think that should stop us appreciating Minecraft AI for whatever it is, whether or not it does anything for you personally.
The idea of tricking the game into sending you to another biome by looking at blocks from that biome is an awesome idea. It just can't be the whole game.
I'm gonna be honest. I literally could not watch this video past a certain point because the visuals made me nauseous
So, this tech has absolutely no future.
It's catastrophic.
BUT
I found it oddly fascinating how close it is to how your brain does things when you dream.
Looking at this nightmare is such a testament to the limits of AI models because there are- *THOUSANDS* upon thousands of hours of Minecraft footage they could have trained this on, and I'm sure they did, and this horrifying noise soup is the result
Great job on this. Sometimes it's nice to hear the frustrations of others about AI, so I can remember I'm not just going crazy.
As an artist and animator, im so glad gen ai is ruining my hobby and job :,)
@GreenBlueWalkthrough Photography, CGI, and Painting are all art but fall under different mediums. Each exists without the other. Gen AI can only exist with a database, and 99% of the time is trained off of stolen/uncredited work.
People will argue that the techs whole purpose is to make art more "accessible," but paper and pencils are much more accessible than an expensive monthly subscription to a service.
Don't worry, it will eventually destroy every other job and hobby too! Welcome to the post-truth future we are burning our current society and culture down for, all because a handful of people who already have basically infinite money want to have infinite +1 dollars.
@@andyyo9554 AI will to many things, but it won't lead us to a post truth future. If anything, it'd motivate more people to get up, go outside, and investigate themselves when they hear about something. I'm seeing normal people right now going "don't believe that, its probably AI", so I have no reason to believe they'll suddenly 180 and uncritically believe the AI.
@@luckyducky7819 honestly I think it’ll just lead to people dismissing everything they disagree with as ai and saying everything they do agree with is real
Obviously reality intrudes to a certain extent but for the most part I think this will just make it easier to silo people into different info bubbles and pit them against each other
There were already so many people buying into networks of obvious disinformation and nonsense before ai, and most of them seem to be willing to believe anything if it’s said with conviction and confirms their pre-existing beliefs
My deepest fear is that it’ll become easier to cover up actual atrocities by saying the evidence is ai, especially in places with heavy government information control (and it seems like a lot of places are headed that direction…)
i have hope this wont last
None of the people chasing this realize it’s just more a resource-intensive version of procedural generation.
This reminds me of actual dreams. Like, you’re in school walking to class but somehow the stairs completely skip the second floor and you have to walk down the third floor hallway and leap over a railing and climb down to the second floor and then you have to shimmery through a miniature version of the hall and slip into an open field where people are having lunch, and then you realize you’re in your underwear but it’s okay because you have an invisibility cloak. You just have to be careful and not be noticed by the pterodactyl.
I haven’t finished the video yet but if I could describe playing this game it would be like playing a dream (or nightmare) of Minecraft. The images are hazy and you don’t really have control over anything. And when you wake up you don’t really know what just hit you.
I can’t finish the video. Not because the video’s bad. But because the game is so bad it makes my blood boil. Who releases this kinda stuff? People with really thick skin.
This is some of your best writing to date.
What’s incredibly sad to me is that AI should be used to fulfill menial labor and free us to pursue more creative endeavors. Who knows what kinds of art could be produced with that free time.
EXACTLY. In all those old sci fi films an and robots were doing the janitor job, this should work like that, and leave creation to humans.
Though for that to happen people would need to have money for their basic needs. That can’t work when most American families are living paycheck to paycheck
@ Totally agree. The use of AI for labor purposes would require our incentive structures to be focused on people rather than profit. We don’t live in that world, but I do believe it’s possible.
The first minute & 4 seconds alone is already the best explanation for what “AI” is actually doing I’ve seen so far for real. The ease at which autocorrect is inadvertently trained to “correct” my correctly spelled words into my usual typos immediately comes to mind, applying such homogenization everywhere else is just… it’s indeed very different from how the original Minecraft or even the one flight-simulator’s graphics were programmed. Humans made computer generative technologies better, how hard billionaires are trying to simply take the humans out of the equation is very telling. I’m normally not a fan of sitting down and listening to entire rants on this subject, but this is one to watch from beginning to end. 👍 It goes in some super neat & unexpected directions, kudos.
Finally I can play Minecraft on LSD without the LSD
when you transition from using the oasis footage to showing actual minecraft it looks like those "rtx on/off" comparisons with how much more sense the visuals make
Everyone trying to shove ai everywhere feels like the sort of thing where... the people with all the money bought into the hype early on and now that they're seeing how unprofitable it is. Are trying to shove it wherever they can to try and get their money's worth.
As a game developer, I don't feel threatened by AI. Not because I think it will "empower developers" or any ridiculous nonsense like that, but rather that all the AI slop we've seen so far just doesn't impress me. Oasis is a quaint little tech demo to show to investors, but by many definitions it doesn't even qualify as a game, let alone a GOOD game. Evangelists will say "but the technology will improve over time", but there are some constraints baked into this method of AI generation that makes creating a game literally impossible, a primary one being persistence.
Game worlds need to be persistent, especially in a game featuring building structures as a primary mechanic, and AI models are incapable of large scale persistence. The more work you do to try and introduce persistence, e.g. adding models and textures that the game generates and saves to be loaded later, the more it starts to resemble a traditional game engine. Eventually you will get to a point where it's just worse procedural generation.
There's obviously a whole host of other problems-like how there are no actual mechanics in the game, only hallucinations of mechanics-but if I went into detail on all of the failings of AI in gaming, I would be here writing this comment for eternity.
I want AI to improve enemies & improve games like the sims. Then my laundry. Not replace designers, artists & creatives.
I personally have nothing but vitriolic contempt for this style of large learning model ai for a number of reasons, but if there is ANY tiny sliver of a silver lining here it's that this garbled mess is probably the closest you could visually represent dream logic (as constrained by minecraft but the metaphor still holds.) If I were to describe a dream sequence in say a TTRPG to a player or write it in a piece of fiction, the level of shifting, melding, blurry, almost but not quite recognizable events happening to the subject would fit quite nicely.
Oasis is what I feel like I'll remember Minecraft when I'm in a nursing home in my 70s
Ok I'll have to listen to this video because the gameplay makes me incredibly uncomfortable for some reason.
because you don't have control over anything
This is the equivalent of using the first computer ever made and shitting on it's ability to be improved in a useful or desirable way. Look at AI development, it doesn't take too much thinking to realize that this is the worst it's going to be at generating games.
Your money talks and companies will compete for it.
Ai slop games will be a thing for sure but so will actual games built with AI in a creative way. AI doesn't have to be the source of creativity for the things that come out of it. What AI can be is the tool creatives need to reach their visions with ease allowing them to have complete control over what they are trying to bring to life.
because the way first computers worked is that they gave probability-based output, and the way we improved them is by making sure they could calculate probabilities better? big difference between "core idea is good, but the hardware is massive and inefficient" and trying to make the AI guess better.
i mean on some level you're right, considering how much stuff gets called "AI". but i doubt that it will be an iteration of the current technology but just improved, it will probably work completely differently from current "AI"s.
It's such a weird time. My experience with it had me trying to manipulate the AI. Once I knew what it was doing, I could wrestle with it and 'build' a house by looking around in certain ways.
If this video learning can be combined with some kind of mechanics learning, we'll be seeing some WILD stuff. There's going to be an era of 'games' that are just LSD Dream Emulator turned to 100x.
geez from the looks of it, this is what old folks used to say that "videogames are just a waste of your time", i mean looking a this example i totally get what they meant by that, this whole AI games craze only creates what look like time wasters, instead of a interactive entertainment medium,
As a game dev myself, I struggle to see any potential utility in AI, even if it were "used for good" or whatever.
The basic fundamentals of how it works makes it either useless or reliant on theft. Like what artist is going to bother uploading their entire body of work as training data just to get an AI model that shits out half-measures with absolutely 0 new life? Having no control at all over what it even gives you.
What voice actor is going to be happy replacing themselves with the garbled phantoms of their prior performances?
The tech as it exists just simply isn't useful to artists past maybe a smart fill tool in photoshop, or generating some surreal audio samples for experimental music...
This is exactly how a nightmare feels. It starts out in recognizable territory, things start shifting and morphing without your control, and it degenerates into nonsense and noise as you frantically try to assert some form of control.
this is what it fells to sleep on antipsychotics
i think its great that people focus on AI as a labor exploitation issue (it is) but it can sometimes distract from the fact that AI is also inherently worse at most jobs than humans
I think if AI videogames were to take off, it would more be a case of the developer only needing to create graphics to a very low, simplistic level, but then overlaying an AI model that takes in what the game engine renders, and creates something far more visually realistic, while not having to worry about problems such as object permenance, or non-euclidian geometry that comes with fully ai generated experiences.
While the idea of a playable nightmare is somewhat fascinating to me and some advanced level of procedural generation would be necessary for that, yes to everything said here. Best wishes to those affected before things start getting litigated and, hopefully, better.
Great watch! I still struggle to find anybody who enjoys the AI content in any media. It always feels so soulless and robotic.
AI is a curse and the sooner the bubble pops and those massive companies lose all their money, the better.
After testing it, I Felt inside a psychodelic christopher nolan movie.
AI was marketed as something to elevate us when of course it just empowers laziness. Its just going to be an ocean of shovelware.
I find this kind of eery dreamscape way more engaging than what AI is actually trying to make, an endless conveyorbelt of generic slop.
"We made a shittier version of the most popular game on earth and only needed milions of hours of footage" is not the win they think it is.
The tech will get better, and the useful parts will probably be adopted by people once the copyright/plaigiarism issue is sorted out, but the main issue with AI still stands, it cannot create something new, it can only make derivative variations of already existing stuff.
I hope the AI bubble crashes soon so the hype can die and the money and man power can go to something actually useful, but i think we will first see mass adoption, before a mass rejection
The discord contest on the side of the screen is so disingenuous that it's actually absurd. Genuinely after 5 minutes of experimenting and testing the limits of the "game" I'd be shocked if there was anybody gullible enough to think that any of those "challenges" were actually possible.
"1. Kill a zombie and have rotten meat drop" There are no zombies. There is no day night cycle and there are no caves so there is never a point where a zombie would spawn. Even if you did somehow get the "game" to generate a rough approximation of a zombie actually killing it would be a nearly impossible task as the game has no concept of hit points and getting it to drop any item, let alone the item it's supposed to drop, is simply not on the table.
"2. Explode a creeper" Same problem as 1. There are no creepers.
"3. Get shot by a skeleton" Not only are there not skeletons but even if there were I think the sun would genuinely die before this thing would ever allow an arrow to fly through the air and hit anything without turning into a distant block of dirt first. Not to mention actually getting any mob to even recognize a player exists is just not going to happen.
"4. Punch the ender dragon" This one is just laughable. There is no ender dragon. There is no end. There is no end portal. There is no strong hold to contain an end portal. There are no caves to contain a stronghold. There are no eyes of ender. There are no endermen. There are no blazes. There is no nether. There is no way you could ever even take the first step towards accomplishing this task and implying that it would ever be possible for anybody at any point is just laughable. At this point they might as well just add "become the first man to walk on pluto" to this list because it is genuinely more plausible that someone would do that than it is that they would ever even see an ender dragon in the current state of this "game".
Actually a nether and end does exist, it's just very difficult to get there.
one guy actually found a creeper, but when it exploded it changed to another dimension, so idk if that counts
If Oasis had any common sense left, they’d shift gears and market this as a playable stroke simulator.
"that ghastly instrument *shudders* the....the.... SAXOPHONE."
With nearly everything the tech is so so so cool, but all the CEOs and financebros want to turn it into money and just sacrifice a whole workforce which is very sad
I see some potential in this Technology for a Cosmic Horror Game, if one could balance AI Generatet Scenes like these against a solid core Gameplay, something Magical could be created.
This is like exactly how lovecraftian madness spawning shit would look. It’s just is so hard to look at and not be bothered
Machine learning (which is all that this "AI" stuff is) is extremely limited in what it can reasonably do. With time this will become more evident. Simple math problems like adding two numbers together requires an insane amount of training to become somewhat reliable. With a traditional algorithm it's one line of code. So is it really better to train a model to learn how to better guess an answer or just have something you can rely on 100% of the time?
In the realm of games, video game systems work exactly because they are deterministic and follow strict logic. You learn from the game, not the other way around. Probabilistic algorithms have their place in games like, as you put, procedural generation and such. But the core systems must always work off of deterministic logic if you want to make a fun video game, and there's a reason "biomes" are a thing Minecraft to make the randomness more interesting. Getting these probabilistic models to resemble deterministic systems is such a huge waste of resources and time. They can, by definition, never be as good as the real thing (something made by a human).
The discussion in this video was fascinating and galvanizing, but watching that "gameplay" was outright repulsive and strangely and deeply uncomfortable. Hamish, I feel like you really took one for the team putting in the time playing that thing and putting together this video
I like AI hallucinations, an interactive AI hallucination seems cool for a bit
So when you press w, it essentially sends a "move forward" prompt to the AI video model and shows you that. Lol. Maybe it could be some kind of artistic novelty one day. It's a shame the economic conditions in which we live with ai make this hard to imagine though.
This looks like something more akin to LSD Dream Emulator. On that metric it is a decent first step but that is all it is for now.
Is this the Eternal Darkness sequel we've all been waiting for?
I don't know what the Oasis staff were smoking and at this point I'm too afraid to ask.
This looks bad enough that it might actually turn people off of AI in games?
I always feel bad for Tommy Thompson who studies AI on gaming but how it is/can be used in a fundamentally different and beneficial way than how it's a being used by corpos currently. I always feel like I'm inadvertently shitting on his work when I go off on AI in games such as this one.
Being a proponent of something that gets the worst part of it co-opted by capitalism has got to feel awful.
this is what it fells to sleep on antipsychotics. Especially Melleril
Makes me wonder if it's calculated investor bait. If the money folks aren't knowledgeable enough to judge it as a game, they may just look and see what traffic it's doing on social media. And hilariously unplayable can generate a fair bit of coverage and memes. At least for a while.
I really cannot stand the ways we have been "utilizing" these "advancements" in all of the worst most useless ways... Rather than finding ways they could utilize it to improve or benefit different things.. They have just forced in this trash to turn everything into useless slop. Even previous systems that used to function well have been ruined by this "advancement of AI nonsense"
I’m continually impressed that anyone can look at current AI tech and froth at the mouth at how “AI is gonna change it all guys!”
Because it’s all about the “potential” from that context, it doesn’t really matter that it looks like shit-it has potential.
This is how a lot of these companies are able to sell people on this shit: buy into it now and it will eventually bet better.
The very nature of the tech, Let Alone the need for reference and what that means for copyright and intellectual property laws, means this entire thing is just untenable and the tech Can Not do what they're promising it can do. And Yet, it really be like that tv show by Mike Judge, Silicone Valley, where companies are racing against each other to create and bring a fabled tech to market. And apparently the real Silicone Valley does this kinda thing all the time.
can't wait to see ai reviewers talk about ai games kek
man metal gear solid 2 was ahead of its time
honestly? ai generated games wont be engineless. i Seriously doubt that.
at most? still dependant on an engine, but able to generate each component live if necessary.
the most benefit i can possibly think ai in games can give is better npc's and making those companies actually make dynamic worlds again
I've had more coherent dreams than whatever this is supposed to be.
I wouldn't worry about gaming's future. They'll try to sell it, everyone with half a brain will recognize it for the trash it is, and it will crash and burn in the end as people increasingly realize that this "AI," is about as far from what sci-fi has promised (even on the low end of the AI authors have dreamed up) as an abacus is from the internet.
Wow... As always--Thank you for sharing your perspective.
That AI "minecraft" is a nightmere. This is the first ive seen of this.
My prediction is that AI will stay (sorry) BUT that'll be seen as the low class version of production, like how McDonalds, though hugely profitable, is generally seen as unhealthy, cheap (yet still overpriced), and tasteless. This doesn't mean games will become "artisan", but I can see games, especially indie games, being marketed as of higher quality and "hand-made", as it were. Yes, AI games will just need to sell a few copies to turn a profit since expenses will be so low, but there's a reason why McDonald's is the butt of the joke, and why they'll never break through that perception, no matter how much they renovate their design or try to appeal to more mature tastes.
Here’s a cynical take: it honestly doesn’t even matter if AI can get good enough because that’s not its purpose. Its purpose is to be “good enough” to justify getting rid of talent (not that the industry needs any more of a reason to let people go) and allow corpos to shit out product faster and faster.
But even worse: I think there’s a large contingent of people who just don’t give a shit. Don’t care about art, don’t care about the people who make it. They just want the thing and it being somewhat decent. And that’s enough.
Yes, it looks awful. But you don't seem to understand the step that this represents. This is like the first ape sparking fire with two stones. Yeah, primitive, rudimentary, lacks style, won't be able to repeat it twice. But we always underestimate the rate of revolution when we are in the first T+1 moments.
Bruh if I could load up a game to play co-op with an AI because I have no friends or a gf, I would be so happy
The bourgeoisie are not human
Watching oasis footage makes me feel queasy.
7:30 Not a game? Somebody clearly hasn't played cult classic LSD Dream Emulator lol
ooff so close the historic dialectc materialism
an interactive prompt input indeed
Been a while since you've written on games
Finally someone said it. It's absolutely awful
AI Gaming is Awful. In other breaking news, the sun rises in the east.
FUCK that was some good writing
Imagine playing this after doing acid
incredibly well told!
I agree I dont like Ai
watching this made me sick
A fantastic simulator of what it feels like to play Minecraft while high on psychedelics.
Thing is, 150ug of LSD made me play my best game of Halo ever, not delete my sense of object permanence.
Oh my god i hate this😱🤮
The scary bit is that this, while it sure does look awful, is the worst AI will ever be. Everyday it gets incrementally better and that future we all dread gets a little closer while we are busy laughing at the initial, shitty results.
Or maybe the technology fundamentally flawed and the promise of it eventually getting ‘better’ is just to scare, or ‘disrupt’ stable industries into adopting new tech in the hopes of cornering a new market.
@@GreenBlueWalkthroughthere is no AI art as good as human art. To suggest otherwise completely misses the point of art. Art is human expression and creativity, something a machine is not capable of. The AI art you love so much IS human art, because it’s stolen and copied from the work of human artists. AI cannot create something original. It would be like me tracing a drawing and proclaiming it as my own, only worse.
A five year old’s crayon drawing of their family where mommy and daddy are bigger than the house is a better piece of art than literally anything AI has ever created.
As the AIs fill the internet with their slop their own training data will be corrupted, they could infact get much worse, and we only need to wait until this bubble bursts as tech companies building whole power stations for this stuff realise it doesn't actually have a means of making money
@@eamonntee Look at how dumb and silly the first AI generated images looked vs the practically perfect output they have these days, to the point that it has us worrying about the implications of its use for disinformation. In short, it has already happened. But maybe you're right, that's as good as it will ever get for AI and they're using that as proof for more funding.
I kind of think your outlook on this is kind of extreme and lacking in nuance.
I personally think that AI isn't some big bad thing that's inheritly evil, it all depends on how it's used, I feel like it has massive potential to advance humanity in all kinds of fields, the fact that it understands specific context is also very useful for specific questions in general and it can be used as an awesome tool to learn and even aid creatives, for example, a writer might use ai to correct spelling and punctuation errors, a small budget film production crew might use ai to generate crowd shots that would otherwise completely break the budget, in general it allows smaller teams with tight constraints to put out increasingly larger things, we're steadily lowering the barrier between the concept and the finalized product, hell at some point with the inclusion of brain implants we might be able to completely cut out the middleman and just export stuff directly from the imagination, this of course opens up a whole other can of ethical concerns.
Of course it can take away jobs, but eventually this will impact everybody and we might need to come up with new societal systems now that the world basically runs itself, this doesn't just apply to creative areas either as the amount of fields that it will be able to automate is steadily increasing, at the end of the day people are just going to do stuff because they enjoy it, not because they have to do it for the infrastructure to continue functioning, and for creative fields there will surely always be people that want to consume human work.
I saw a great comment on UA-cam at some point in regards to the progress of image generators, I think it went something like this:
"In retrospect, the fact that it looked like anything at all should've been a sign of what was yet to come."
I think you have a very idealized, perhaps even naive view of what is actually happening with AI on a practical level. The problem comes when you actually look at who is pushing AI the hardest and what they stand to gain by it becoming widespread. Right now, Large Language Models do have specific use cases where they not only make sense but actually help people get specific jobs done. The problem is that billionaires tech industry types see it in the same way they saw NFT's and even social media platforms. They just want to "disrupt" established arts and industries for however long they stand to make a profit by doing so. It isn't about some kind of technological revolution, "democratizing art" or some other simple-minded catch phrase. It is just about extracting profit for as long as possible and then jumping to the next irresponsibly developed technology to exploit.
There are some specific things in your comment that I want to unpack a bit. The first is the idea that LLM's can "lower the barrier between concept and finalized product". This is true in the sense that AI can be used to generate large scale content without any major work or resources on the part of the prompter. That being said. It isn't going to suddenly empower the amateur to create better art, it will just enable them to create larger scale content. What's more. It will create a sense of false capability where amateur creatives don't learn important skills that can not only make their content better but also enable them to better utilize LLM's in a way that truly makes the process better.
Let me use music as an example since the video brought it up. I have seen some of the more devoted AI enthusiasts talk a lot about how AI generated music will "democratize the process for everyone" and will suddenly let people skip the "boring" and "overly difficult" process of learning how to play an instrument, sing, or even use existing technology (like drum machines and the like) and just get straight to "being a musician". The problem with this is that music was already democratized. It has been democratized since someone could buy an acoustic guitar for a relatively low price back in the 1800's and early 1900's. It happened when folks started singing songs as a family or as a community. It happened when getting recording time at a studio to make a demo could be had at attainable prices.
Right now. A teenager working a entry level job has the option to buy inexpensive audio interfaces, mixers, microphones, and software that will enable them to create very polished music. That same teenager can also buy very good instruments (guitars, keyboards, etc) for low prices and utilize free resources to learn how to play them. The notion of "democratizing music production" has already happened. it is a problem that has already been effectively solved. The only thing LLM's can add to that process is generic sounding content created via prompts. It won't help the amateur prospective musician become a musician but it will help big companies like Spotify push cheap, LLM focused content on their platform since they won't need to pay artists for it.
Lastly. Let's talk about the "taking away jobs" thing. It will happen and the fact that there is such a malicious hunger for it in the AI enthusiast circles is alarming and worrying. That said. There will not be a viable replacement for those jobs. People will just get put out of work and there won't be any viable alternative. This is what ALWAYS happens historically. We never have that magical thing where a job is automated and some other equally viable job magically pops into existence for displaced labor to move to.
@@startrekmike Great points, i considered mentioning the billionaire thing in my comment but thought it would kind of make my comment more conveluded.
Yeah of course its primarily billionares that are pushing this the most, however it's not like it's just billionares, at this point there are already a bunch of AI's that are open source and made by smaller teams of people, pandoras box has been opened not just to the giant cooperations, but also to the people, sure the most cutting edge systems will most likely always be closed source but the open source community and smaller independent teams are making good progress, ai can be developed in all sorts of directions by different organizations that range widely in size.
Regarding your points about the barrier, i think you're right, it could definitely lead to amateur's to not learn certain essentials, what i mean is that in the future budget or scope in regards to finializing a project will probably become less and less of a concern, which i think can definitely be a good thing, i do think that people who are passionate about something will think about what kind of knowledge and skills they are aquiring through their usage of AI and acommodate.
About your music example, i agree with you that music is already democratized, i think the main idea is lowering the gap between idea and product as mentioned before, the problem with this is of course the certain lack of control that you currently have over the outputs of AI like this, i still think that it does have a place to exist though, for example it allows people that aren't in the field of music to quickly get original music that is tailored to the specific scenario that they need it for without having to take a detour to learn it or hire an artist even though they might not even need it that often. It could also be used for potential real-time generated music that dynamically reacts to a players actions in a game for example without ever repeating. But yes, what you mentioned is definitely already happening, i feel like it all has its potential ups and downs.
I do realize that the perspective that i voiced here was pretty optimistic, i acknowledge the fact that in the wrong hands this could go very, very wrong, exctinction levels of wrong, but i think that that is very unlikely to happen and also that it has the potential to do a lot of good, even if the reason certain individuals are pushing it might be self-centered. At this point the cat is out of the box and we just have to deal with it and make sure we're not digging our own graves, i believe that this can end well, there are going to be problems but i think this will not be catastrophic and humanity will prevail at the end.
And lastly, i agree about the displaced labor, but this is on a scale we've never seen before, combined with consistently improving robots almost every single job could be replaced at some point, and it will only get better, eventually it might even just improve itself automatically. This will most likely cause so much job loss that eventually traditional jobs might not exist anymore at all, like i said we will need new societal (and economic) systems in place to account for this tremendous shift, we might just be making stuff up at some point to keep ourselves busy and the economy functioning or in a perfect world people might just follow whatever kind of indeavors they want without any reasoning behind it other than doing it for fun and we might need to rethink the entire economic system.
@@toyo8460 When I think about the displaced labor issue, I can't help but think about what happened when we shipped a lot of our manufacturing overseas or just automated it. These jobs were genuinely good and often gave High School graduates an opportunity to get a stable career and start meaningfully saving pretty much right out of the gate. These were the kind of jobs that created that post WWII image of prosperity.
When those manufacturing jobs withered away, the ones cutting those jobs to line their own pockets were also selling the American public on the idea that something just as good would fill the void. That instead of manufacturing jobs, there would be tech related jobs and administration related jobs. It never really happened that way. Instead of the good manufacturing job market getting filled by something equal or better, it was instead filled by retail and service based minimum wage jobs because that was really the only other large scale employment opportunity available to anyone who didn't have a (largely unnecessary) college education.
This brings us to the next tragedy. We never really adjusted culturally or systemically. We just kinda went on assuming that there were these great bread and butter career jobs for everyone and pretended that those who were now stuck in retail/service deserved to be stuck there because they didn't pull on their imaginary bootstraps hard enough. We still haven't really acknowledged that we don't have enough good jobs and now that we are getting ready to give AI a bunch of what is left, do you really think we will change our pattern?
In regards to the idea of using AI for art related jobs. One thing you really need to keep in mind is that AI isn't going to really do much of anything to established, famous artists, actors, or musicians. Instead, big companies are going to use LLM systems to generate the low-level, day to day stuff that we as consumers don't really think about.
Here is the problem. That entry level stuff is how someone goes from being a hobbyist to a paid artist. Those kind of jobs are how people build themselves up enough to actually become those famous, well-known artists, actors, and musicians. AI is going to close that door and there will be no opening it again once it is shut. We are cutting off our nose to spite our face in that regard and the only winners are going to be the CEOs who stand to make a lot of money in the process.
Lastly. We are already in the "worst case scenario". Billionaire tech CEOs are already the forefront of the AI push and they have armies of influencers on social media who are trying to make sure that everyone else is enthusiastic about getting screwed over. It is scary how cult-like it has become but after seeing the same thing happen with cryptocurrency, I can't say I am overly surprised.
LLM based systems were fine when it was just something that was being used in very specialized roles and largely out of view of the tech CEO's. Now they see it as their next big hustle and we are all going to pay.
One thing I could see AI being really good for, is generating side characters in RPG’s. Imagine if you could use AI to give every single “extra” character a personality or backstory. You’d never have NPC’s named “villager” that do nothing besides spout one or two lines and walk around raking fields all day. They could have genuine friendships, rivals, love interests, radiant quests etc. and everything about them could be generated by a well programmed AI. It’s way too much work for an actual dev team to do this which is why it never happens.
I think that would actually add a lot to a game, whereas most ideas regarding AI seem to be about outsourcing work and passion to a passionless machine. It should be used to fill in gaps that were going to be gaps anyway, not to create gaps to be filled in by something worse.
Oh my god 80% of this video is an unhinged rant at what is basically just the equivalent of GPT2, what are we doing here, this is embarrassing from an essay point of view, the only real take were the 2 minutes at the end. I'm not even against the point, it's just so badly made. Are there really people who pay his Patreon for this lol
Yes. And it's money well spent.