#kinda predicted nfts there | Explore Tumblr Posts and Blogs

eleemosynecdoche · 1 year

Text

A provisional reading of TH19/UDoaLG:

So the incident that's happening is that the effects of TH18 and 18.5, the games that were kind of about NFTs, have caused Gensokyo to lose its metaphysical state of ownership- it's disrupted the space, you might say, and the animal yakuza are trying to muscle in and claim Gensokyo for themselves.

Another way of putting it is that LLM art and text generation has arrived in Gensokyo, and that makes Gensokyo vulnerable to being captured through legalistic shenanigans.

To counter this, Zanmu, a metaphor for an actual AI in the scifi sense, who is convinced she's perfectly rational, decides to keep Gensokyo "free", but ultimately subordinated to her dominance. (I think you can really go hard on Hell as America with that one, Zanmu as an incarnation of neocolonialism.)

To counter-counter this, Suika sends Reimu to short-circuit the planned network of clues and then lose. Rather than pick a fight that really can't be won, Zanmu is forced to admit that she can't predict everything, and that people will behave irrationally and against their "best interest". (600-year-old monk masquerading as an oni forced to admit the realist theory of international relations is inadequate after shrine maiden loses fight with her!!)

This is then matched by Hisami realizing that she can't make Zanmu care about her by being efficient, and so she acts contrary to Zanmu's wishes. (And of course, because she's crushing on Zanmu, Zanmu can't predict her irrational behavior driven by love or infatuation, because Zanmu is an AI metaphor.)

And then Zanmu recognizes Reimu as like a younger her. Someone she can relate to as a person. There's hope for her/America yet! And the overall thesis, of imperfection being good, being a thing to find joy in, is, I kinda think, reinforced by how failgirl everyone is.

Anyways, I think interactions between Zanmu and Hecatia would be fun and interesting because of Hecatia standing in as a kind of metaphorical good representation of chaos/America. I also think it's really funny that ZUN flips the SMT1 identifications as far as Law and Chaos go.

#th19 spoilers #touhou meta #touhou

4 notes · View notes

thepcdos · 3 years

Text

Please, Cyberpunk wasn't supposed to be a How-to guide!

#kinda predicted nfts there #didnt you bill

8 notes · View notes

nostalgebraist · 2 years

Text

IMO, generative ML models are overhyped right now.

(Generative ML models are things like GPT, PaLM, DALLE-2, Imagen, etc. -- the models that make what people call “AI generated” content.)

I find myself almost wanting to “short” them, in the financial sense. Maybe I should? Like, on prediction markets, or just in personal bets.

----

These models tend to have a large “impressiveness-to-usefulness gap.”

Remember the GPT-2 staged release? OpenAI claimed they were worried about people using it to automatically generate disinformation, and things like that. And other people worried about this too.

But when it finally did get released, even the scary full GPT-2 just . . . wasn’t very useful for anything?

It was really cool, and did get used for lots of internet art-and/or-comedy projects.

But those are exceptions that prove the rule, because they always hinge on “hey, look at this thing that an ‘AI’ made!” If you stop caring whether the text was machine-generated or not, and just ask whether it works on its own for some practical purpose -- then the magic disappears.

Likewise for GPT-3. For the amount of buzz it got when it came out, it’s pretty striking how little effect it’s had on the world in the two years since then. There are a bunch of GPT-3 startups, but unless you use one of their relatively niche products (you probably don’t), none of the text you read in your everyday life was written by GPT-3. Except the stuff where the point is that it was written by GPT-3.

(And even some of those startups are doing the “hey, an AI made this!” approach, like AI Dungeon. In OpenAI’s 2021 blog post about GPT-3 apps, one of the two examples they chose to highlight was Fable Studio, a company making AI for VR characters -- an application where “wow, this AI is so advanced!” can do a lot of the work.

...and where is Fable Studio now? They’ve, uh, pivoted to NFTs. Specifically AI for NFTs, and more specifically a behavioral, non-linguistic kind of AI inspired by The Sims. “GPT-3 is frustratingly useless” is a core part of their pitch for the new project.)

And now people are freaking out about the idea that DALLE-2 will replace human artists, and it just feels like the same story again.

The pictures are pretty. They’re technically competent by human standards, as GPT-3 output often is. Yet they’re rarely actually good by human standards: I haven’t seen a single one that would make me click “follow” if it had been posted by a human artist on tumblr.

And as with GPT-3, it’s virtually impossible to tell the system what to create in the fine-grained way you would expect when collaborating with another human. Mostly, it just makes a competent version of whatever thing it randomly happens to end up making.

----

Most of the impressiveness-to-usefulness gap is about that last point. Collaboration is crucial in creative work, and these models can only collaborate with us in a very primitive way.

You can’t impose a house style on them. You can’t tell them “this is a good start, now here’s what I want in the next draft.” (Although you kinda can with GPT-3 now, but that raises the meta version of the issue -- can you rely on it to respond to your feedback in a predictable way?)

You can’t ask them to make 500 different video game objects that all look the way the objects in your game are supposed to look. They can only make 500 objects that look convincingly like they’re from some video game.

If models like DALLE-2 could be told “this is a great art style, now hold it fixed,” and then told to generate many pictures in the same style, then IMO they would be much closer to doing the work of human artists.

But it’s hard to do this kind of thing with generative models without spoiling their magic.

The models get their magic by learning from immense amounts of data, mostly scraped from the internet. What they are good at is imitating the structure of information that already exists ubiquitously on the internet.

When we “tell them to do things,” what we’re really doing is filling in a piece of some kind of information-structure that can be found all over the internet. When you prompt DALLE-2 or Imagen, you’re saying something like “make a picture that would have this alt-text,” or “make a picture that would have this caption on Instagram.”

The reason you get such coarse-grained control is because real stuff-on-the-internet is only coarsely predictable from its constituent pieces.

And what’s on the internet is simply what it is. You can’t just go out and tell millions of Instagrammers to write captions that are more informative in some specific way. If people are making the data with you in mind, the data isn’t big enough to be relevant.

So with something like varying the style and content independently, there are 2 basic approaches:

- Reify “style” as some specific thing about the image that can be computed automatically, as in the style transfer literature. This works reliably, but loses the magic: you can control “style” very precisely, but at the cost of using a limited, hardcoded notion of “style” that can’t make itself less limited by learning from massive data.

- Use the model’s innate conditioning mechanism, adding “in the style of Salvidor Dali” or whatever to your prompt. This gets you a much more nuanced, impressive concept of “style” that gets only more so with more data . . . except that you can’t put anything in that wasn’t in the data to begin with. You get the information-structures that the Instagrammers (et. al.) gave you, and you get only that, and no more.

I think some narrow versions of the problem will get resolved over time, like the one that is the topic of the Scott-Vitor bet.

But the “Instagram bound,” so to speak, seems like a fundamental limit. We might overcome it, but only by doing something fundamentally different from anything you’ve seen yet.

#ai tag

221 notes · View notes

juanitasupreme · 2 years

Note

chungha missing for real what happened to her 💀💀 tbf for sunmi from what I'm seeing she's been given the kiss of death on twt for the nft stuff. no matter what she does now she will get shit for it. A bit sad if im honest cause she's actually always been a very likeable person. it is kinda funny tho to see all the stans who came for her bend their asses over trying to now defend their own faves for their nft activities as u predicted a couple months ago. twt will always be twt

People are being hypocrytes but again liking someone then throwing them in the mud after some times is not new. People will come to there senses hopefully

#anon #sunmi

13 notes · View notes

cadenzaaa · 3 years

Text

Quick post-covid special notes (haven’t done one of these in years). I’m assuming I don’t need to tag this because 1.) I’m 1000 years late and 2.) This hasn’t been a SP blog in ages, and I don’t think I have many invested SP fans still following (but also hi, any sp mutuals who may see this). I managed to go into this mostly unspoiled, and I’m proud of myself.

- Creek was one of the first ships I ever shipped at like, age 10, so I’ll never forget them tbh. I’m sooo happy they seemed to have made it on both timelines. FUck Cartman for giving Tweek a panic attack though.

- This may be colored by how rusty I am on my South Park canon, but I love how M & T managed to weave past covid specials into this one. Additionally, I love that on top of the careful arrangement of plot points, the originality etc, they still bring it back to the familiar, simple, hilarious story line of Cartman plotting to kill Kyle.

- Butters was great. I had my predictions of what he would be like, but they managed to exceed my expectations in a way that totally sent me. It might be because I’ve been involved in fan communities who wholeheartedly HATE NFTs, but the jokes landed perfectly for me, it’s such a Butters thing. I wasn’t expecting NFT jokes, but this was a perfect use of them,

- We saw Ike!! Also loved the brief parts for Kevin Stoley.

- Cartman’s future really was contingent on Stan and Kyle’s friendship surviving. Cartman really had a change of heart and gave up his future, leaving it to fate, growing from the usual “Cartman wants to kill Kyle” trope. And they were all happy but him in the end. This is a lot for me to process tbh. I don’t like Cartman, but there was character development here, I think.

- Loved the therapy perspective in Stan’s final speech - “We can’t control covid, but we can control how we respond to it.” I’m paraphrasing here, but this is such good advice. It may seem obvious, but we often get caught up trying to change things beyond our control.

- Clyde cracks me up. I’m kinda sad about him, but also how did he not gaf about Tweek and Craig? Wtf, Clyde? This also somehow makes 100% sense to me.

- Stan is a better person than me for apology to Randy.

12 notes · View notes