#i mean i KNOW my posts arent ai generated but just the thought that they come across that way lol | Explore Tumblr posts and blogs

theothin · 1 year ago

Text

@lilietsblog continuing from this post in order to not have this conversation buried under the other one before it.

so my position on "AI" art is - video game opponents have a better claim of being an "AI" than the image generating machine, much like the text generating machine. at least they are trying to imitate what a real AI would do, internally. at least they are reacting to circumstances based on a model of the (simplified digital) world;

What do you mean by "what a real AI would do"? Real AIs have existed for decades, with all sorts of levels of complexity. Most of them haven't gotten this kind of pushback about the label.

As for an internal model of the world, current LLMs do have those. They can, for example, form records identifying the United States as a country bordered by Canada to the north, Mexico to the south, the Atlantic Ocean to the east, and the Pacific Ocean to the west, consisting of 50 states, 3.8 million square miles of land area, and a population of 333 million. They can form these records about anything, and draw conclusions about actions to take based on those records - and some of those records are wrong, but that's no different from humans.

- the "art" part is giving the machine the prompt, not what the machine does. any more than photoshop is the one doing art when you add a gradient or a filter to your picture. just because the process is obscured and randomized doesnt make it any less mechanically programmed;

I think this is mostly a philosophical question. Suppose I type out a paragraph outlining an image I want and put that paragraph into an AI art generator, and I also send that exact same paragraph to a human artist as a commission. Am I more or less of an artist of one of the resulting pieces than the other? Does it matter?

Of course, commissions are usually iterative - and the creation of AI art can be, as well. I think a lot of people think of "AI art" as referring specifically to the process of inputting a sentence or paragraph and getting an image, but LLMs are complex enough to support all sorts of more complicated processes.

Personally, a type of AI art generator I'd love to use is one that involves a chat interface that allows you to give feedback based on previous images it has generated and to use them as reference points for future ones. For example, it might generate an image of a character you requested, and you can say, "keep these aspects but change these other ones". Then once you get something you're happy with, you can tell it to classify the image as an example of how that character is defined for the purposes of the project in question, to keep it as a reference point for future images. Ideally with the ability to save local records of your interactions with the AI in question, so you can reopen it later and maintain progress on the project, rather than needing to maintain a consistent session.

I think there's good odds that someone has already made a tool like this, so if anyone reading this knows of one, please let me know!

- like, art is communication. the image generator isn't communicating anything to you, it doens't have an internality to communicate. the prompter does. and they're not communicating with the AI, they're communicating with the viewer. like using text to speech or something like that. a technological intermediary;

Sure it is. It's communicating its understanding of the subject matter, in terms of the parts of the world it has studied and the resulting thought-code it has written into itself. Which is also a thing a human might do, through the exact same process.

- art is NOT in fact universally accessible. "i am disabled and I do art and teach art" is not an argument any more than "well I am disabled and I don't need a wheelchair" is an argument against wheelchairs. there are different disabilities??? i mean like everyone can do SOME form of art but they arent fucking. mutually substitutable. not everyone can draw a picture and i dont mean in a skill way. someone could have been a fantastic artist and then lost the ability bc their hands got fucked up;

Well said.

- speaking of mutual substitution, machine generated art is not the same medium as traditional and digital art any more than those are the same medium as photography. the means of obtaining the result are categorically different, come with categorically different challenges and skill requirements. you wont judge a photo on how photorealistic it is and you wont judge a painting on how difficult it must have been to stage. machine generated art is a third thing and must be clearly labeled in any situation where the difference between a photo and a painting too would be relevant;

No, medium is the material an art piece is made of. Painting, photography, and digital art are distinct mediums. Human-drawn digital art and machine-generated digital art are two techniques within the medium of digital art.

Technique can be relevant, but it's a smaller difference than the difference between a painting and a photograph, and mixing techniques is more straightforward than mixing media. A piece of digital art can be 1% human-drawn and 99% machine-generated, 99% human-drawn and 1% machine-generated, or anywhere in between.

If a human uses an AI to generate an image, then draws their own digital image using the AI-generated image as a reference, should they have to describe their drawing as machine-generated? Does this change if tracing is involved?

This is another situation where I think it's best to not find ourselves in the position of having to ask the question in the first place.

- aaand the means by which it is obtained are sketchy as hell. in a perfect future utopian UBI reality it maybe wouldnt be a problem but in the current now reality it hella is. and regulation based on existing flawed legal frameworks is a better solution than "waiting for the underlying issue to be solved" much like "homeless shelters" are not a better solution to the problem of homelessness than "housing for everyone" programs but are hell of a lot better than nothing;

What means do you take issue with, and what laws do you propose? I most often see people making these arguments in the context of advocating for using copyright laws against AI, but I was under the impression that was not your position. Has that changed?

And I'm not talking about waiting and hoping. I'm talking about taking existing frustrations around the job market and putting that energy towards UBI advocacy now, rather than wasting it on seeking laws that will cause more harm than good.

- also I think it would still be an issue in the perfect future utopian UBI reality because its a deepfake machine. its just a fact of life for photos but people not wanting the machine to read their art style and learn to replicate it are valid regardless of monetary concerns. what the machine does in remixing images is not in fact the same as a human looking at the picture and reverse engineering the technique by which it was done / having thoughts and ideas about how to do a similar thing. it is ethically and practically different so there you go, youre both wrong and im right. praise me :P

Faked images will be an issue regardless of UBI, and they have been an issue long before they were possible with machine image generation. Meaningfully addressing faked images needs to come from building public awareness and skills in fact-checking, it will never be enough to use methods that depend on the expectation of being able to know the technique used to create them.

It is valid for people to feel whatever desires they happen to have. When those desires are unreasonable, others are not obligated to act in accordance with them. The desire to exclude one's art from machine learning is one such unreasonable request.

When a person looks at an image, their brain writes code corresponding to that image, tags aspects of it with known identification, and uses the resulting code as a possible reference for future actions, including art. That is not a metaphor, it is 100% literal: brains run on computer code that forms reference points based on data recorded from billions of observed images, and they act based on a mix of imitation and extrapolation. It is the exact same process as LLMs use.

@lilietsblog continuing from here:

"the ability to formulate open-ended instructions to itself" does not strike me as proof of intelligence, no, when those instructions result in the "how do you know where the nearest mcdonalds is" kind of interaction. its just another layer in the cake of stupid

I didn't say "proof" of intelligence. I'm saying it is literally what thinking is. If you would define "thinking" another way, what definition would you use?

LLMs reason in different ways than humans do, and that reasoning can lead to odd conclusions. But reaching odd and potentially stupid conclusions is, as we both know well, not proof of an absence of intelligence, just of a different perspective. Those are issues to work on, but they're issues of implementation, not indications of a fundamental lacking.

Even if current LLMs might not yet fully meet the expectations for "human-like" intelligence, I think we can pretty definitively say that they can match or surpass the intelligence of some animals, which is still plenty of reason to describe them as AI. Perhaps the best comparison is birds, which can be plenty smart as animals go. Do you think current LLMs are unable to match the intelligence of a bird?

#gen ai #ai art

20 notes · View notes

synthetic-sonata · 2 years ago

Text

💞 making a quick link to all the important links on my blog and pinning it bc i forgot you can do that on tumblr. on any other site i am knightofpherae / knightofpheraee if you don't recognize my username here, but know me .

💗 mocha / aria | he/they/etc | 19 ( or over )

🍰 homepage / about / F.A.Q / carrd / commissions / oc ask blog / pokemon irl blog / artfight

🐈 robotenjoyer and also ai/cyborg/android enjoyer. terminally insane. ill about my OCs and usually one fictional guy constantly . doeboy or catboy ( gender neutral ) depending on how i feel . i am autistic and my special interest is pokemon and related media . system . aro spectrum

🖌️ my art tag is #aria doodlez and sometimes i tag art i rb as insp, which mostly just means i rlly like ur art and artstyle and that post is smth i wanna look at again later. commissions and art trades are probably not open.

🍒 i am 19 or over by the time of writing, and i reblog nsfw/suggestive things sometimes but 99% of the time its just nsfw textposts/jokes, but i never reblog straight up porn. if i did i'd have a sideblog for that lol . these r all tagged as either nsfw or suggestive, so blacklist those if it makes you uncomfortable.

❓ know what i reblog and blacklist tags accordingly. i do not reblog porn. i often dont tag slurs anymore unless its like, the r slur. i talk or rb kink discussion sometimes and reblog nsfw jokes. you should probably filter those words n tags yourself if you take issue w them. i do not really prevent people from following this blog but just be aware of these things Or else i told you so !

⚡ oh also if you have the default tumblr icon or just an icon of a random woman and no desc ( or a shady desc ) or likes or posts or anything and your acc hasnt been here for long im Going to assume ur a bot and report/block you. If you somehow arent uhhh. Woops. this goes for asks too. i Block annoying people too . Watch out

🎨 as for art credit and such, if you post my art somewhere, do not repost it just link to it. and also do not use my personal art as your icons or whatever ( personal meaning ocs and stuff ) unless i know you and/or give you permission. thankyou.

( art tags, because they're sort of weird: tags under just the character name is for any general thing related to said oc, character name ( oc ) is usually for only the art of the oc or for specification. It's a little messy sometimes )

___

feel free to send me asks, questions, or dms whatever - i love talking about things i like, like my ocs or interests or characters i enjoy! you can also send art requests maybe involving characters i like from things i'm into, but don't expect me to draw those.

don't ask about my banner image i thought it was funny when i was younger and i do not want to change it ( yet )

#aria talkz #about

17 notes · View notes

somnilogical · 5 years ago

Text

modular "ethics":

a wrong and two rights make a right

<<I've been known to cause outrage by suggesting that people who really care about something shouldn't have romantic relationships. Think what would happen if I dared to suggest that those people should also seriously consider getting castrated. That would be crazy! And who am I to suggest that basically everyone claiming to be doing good is faking it? Then people would feel bad about themselves. We can't have that!>>

https://squirrelinhell.blogspot.com/2018/02/men-have-women-are.html

previously i talked about an infohazard about altruism that seemed to fuck with grognor. it feels useful to pass by the dead and look at their lives and choices.

i dont think that castrating yourself is a good intervention for doing stuff you care about, like this is patchwork constraints for an unaligned optimizer. if you arent altruistically aligned from core values, castrating yourself wont make you more aligned.

the "altruists" having babies thing is actual insane and pasek is right about that. pretty much all of society will try and gaslight you about this the way sometimes people are gaslit about "i need to have sex with lots of attractive fems to keep up my moral so i can do super good stuff afterwards.". like if people want to do good for the world it will flow out as a continuous expression of value not some brent dill kind of deal that institutions like CFAR accepted until there was too much social pressure for them to maintain this facade.

the entire premise that morality is this modular thing and you can help set the utility function of an FAI while being a terrible person, is wrong. yet organizations like CFAR keep thinking it will work out for them:

<<We believe that Brent is fundamentally oriented towards helping people grow to be the best versions of themselves. In this way he is aligned with CFAR’s goals and strategy and should be seen as an ally.

In particular, Brent is quite good at breaking out of standard social frames and making use of unconventional techniques and strategies. This includes things that have Chesterton’s fences attached, such as drug use, weird storytelling, etc. A lot of his aesthetic is dark, and this sometimes makes him come across as evil or machiavellian.

Brent also embodies a rare kind of agency and sense of heroic responsibility. This has caused him to take the lead in certain events and be an important community hub and driver. The flip side of this is that because Brent is deeply insecure, he has to constantly fight urges to seize power and protect himself. It often takes costly signalling for him to trust that someone is an ally, and even then it’s shaky.

Brent is a controversial figure, and disliked by many. This has led to him being attacked by many and held to a higher standard than most. In these ways his feelings of insecurity are justified. He also has had a hard life, including a traumatic childhood. Much of the reason people don’t like him comes from a kind of intuition or aesthetic feeling, rather than his actions per se.

Brent’s attraction to women (in the opinion of the council) sometimes interferes with his good judgement. Brent knows that his judgement is sometimes flawed, and has often sought the help of others to check his actions. Whether or not this kind of social binding is successful is not obvious.>>

https://pastebin.com/fzwYfDNq

<<AnnaSalamon 2/6/09, 5:54 AM

Aleksei, I don’t know what you think about the current existential risks situation, but that situation changed me in the direction of your comment. I used to think that to have a good impact on the world, you had to be an intrinsically good person. I used to think that the day to day manner in which I treated the people around me, the details of my motives and self-knowledge, etc. just naturally served as an indicator for the positive impact I did or didn’t have on global goodness.

(It was a dumb thing to think, maintained by an elaborate network of rationalizations that I thought of as virtuous, much the way many people think of their political “beliefs”/clothes as virtuous. My beliefs were also maintained by not bothering to take an actually careful look either at global catastrophic risks or even at the details of e.g. global poverty. But my impression is that it’s fairly common to just suppose that our intuitive moral self-evaluations (or others’ evaluations of how good of people we are) map tolerably well onto actual good consequences.)

Anyhow: now, it looks to me as though most of those “good people”, living intrinsically worthwhile lives, aren’t contributing squat to global goodness compared to what they could contribute if they spent even a small fraction of their time/money on a serious attempt to shut up and multiply. The network of moral intuitions I grew up in is… not exactly worthless; it does help with intrinsically worthwhile lives, and, more to the point, with the details of how to actually build the kinds of reasonable human relationships that you need for parts of the “shut up and multiply”-motivated efforts to work… but, for most people, it’s basically not very connected to how much good they do or don’t do in the world. If you like, this is good news: for a ridiculously small sum of effort (e.g., a $500 donation to SIAI; the earning power of seven ten-thousandths of your life if you earn the US minimum wage), you can do more expected-good than perhaps 99.9% of Earth’s population. (You may be able to do still more expected-good by taking that time and thinking carefully about what most impacts global goodness and whether anyone’s doing it.)>>

https://www.greaterwrong.com/posts/4pov2tL6SEC23wrkq/epilogue-atonement-8-8

like opposing this isnt self-denying moral aestheticism or a signalling game of how good you can look (credibly signalling virtue is actually a good thing, i wish more people did it by for instance demonstrating how they win in a way that wouldnt work if they werent aligned. whose power seeded from their alignment.). its like... the alternative where people do things that it makes no sense for an altruist to do and then say that when they go to their day jobs they are super duper altruistic they swear; compartmentalizing in this way ...doesnt actually work.

people who want to obscure what altruism looks like will claim that this is moving around a social schelling point for who is to be ostracized. and that altruism as a characteristic of a brain isnt a cluster-in-reality that you can talk about. because it will be coopted by malicious actors as a laser to unjustly zap people with. these people are wrong.

both EA and CFAR are premised on some sort of CDT modular morality working. it is actually pretending to do CDT optimization because like with brent at each timestep they are pretending to think "how can we optimize utility moving forward?" (really i suspect they are just straight up mindcontrolled by brent, finding ways to serve their master because they used force and the people at CFAR were bad at decision theory) instead of seeking to be agents such that brent when brents plans to predate on people ran through them, he would model it as more trouble than it was worth and wouldnt do this in the first place.

CFAR and EA will do things like allowing someone to predate on women because they are "insightful" or creating a social reality where people with genetic biases who personally devote massive amounts of time and money to babies who happen to be genetically related to them and then in their day job act "altruistically". as long as it all adds up to net positive, its okay right?

but thats not how it works and structures built off of this are utterly insufficient to bring eutopia to sentient life. in just the same way that "scientists" who when they arent at their day jobs are theists are an utterly insufficient to bring eutopia to sentient life.

<<Maybe we can beat the proverb—be rational in our personal lives, not just our professional lives. We shouldn’t let a mere proverb stop us: “A witty saying proves nothing,” as Voltaire said. Maybe we can do better, if we study enough probability theory to know why the rules work, and enough experimental psychology to see how they apply in real-world cases—if we can learn to look at the water. An ambition like that lacks the comfortable modesty of being able to confess that, outside your specialty, you’re no better than anyone else. But if our theories of rationality don’t generalize to everyday life, we’re doing something wrong. It’s not a different universe inside and outside the laboratory.>>

to save the world it doesnt help to castrate yourself and make extra super sure not to have babies. people's values are already what they are, their choices have already been made. these sort of ad-hoc patches are what wrangling an unaligned agent looks like. and the output of an unaligned agent with a bunch of patches, isnt worth much. would you delegate important tasks to an unaligned AI that was patched up after each time it gave a bad output?

it does mean that if after they know about the world and what they can do, people still say that they specifically should have babies, i mark them as having a kind of damage and route around them.

someone not having babies doesnt automatically mark them as someone id pour optimization energy into expecting it to combine towards good ends. the metrics i use are cryptographically secure from being goodharted. so i can talk openly about traits i use to discern between people without worrying about people reading about this and using it to gum up my epistemics.

26 notes · View notes