Generative AI flooding online crocheting spaces with unrealistic amigurumi pics

https://twitter.com/LauraRbnsn/status/1741494511849578753

91 86 3 years ago

notamy• 3 years ago

https://nitter.net/LauraRbnsn/status/1741494511849578753

OT: Are there any alternate HN front-end that fix this? I don't want to create an account to see the thread. I know there are browser extensions, but for security reasons I only enable ones that are vetted by Mozilla. I think HN used to have a "web" link for paywall bypass, but I don't see it here.

danielbln2• 3 years ago

OP posted a nitter link, which already does everything you want, no?

brucethemoose23• 3 years ago

There are extensions that will "nitterize" twitter links among others, but TBH the one I have been using is now unmaintained and I need to find another one.

justsomehnguy2• 3 years ago

You would be accessing Twitter posts through a 3rd-party service, which would see everything you would see. Because, you know, it can't be any other way?

So your insistence on 'privacy' yet 'only vetted by Mozilla' is misguided at best.

politelemon• 3 years ago

What do the people posting the fakes, get out of it?

ravenstine• 3 years ago

Social validation. When you're narcissistic, legitimacy really doesn't matter so much as others believing you're amazing. And it works because it's hard to prove anything online. Maybe this means hobbyists will actually have to meet in person again. Or online crocheting communities need anti-AI AIs to detect and block impossible fakes.

ericmay• 3 years ago

I think you’re right. Generally speaking things go through cycles of centralization and decentralization.

Generative AI is going to cause centralization of trust back to authorities and authoritative sources for now. It’s already occurring for the news and media, and it will occur for hobbyist communities as well.

Perhaps with a public/private key “is human” verification we might see continued success for creators and content on the Internet, but I think that using AI to detect AI is likely to falter because it’s just easier to meet in person or use a key instead. It’s a matter of category difference.

brucethemoose23• 3 years ago

Yeah. Even sites that encourage "real identities" like Facebook are full of would-be liars seeking validation.

...But anti AI-AIs are not going to work. Even of they work technically (and I strongly suspect they will not), the keepers of the internet do not have a financial incentive to pay for it.

Seems like the larger internet will be zombified, and users will go back to private online circles and meatspace focused interaction. Thats not so bad.

kiba2• 3 years ago

Money.

And maybe internet points.

Unfortunately, there are a lot of people who don't have any pride in their work.

ceejayoz• 3 years ago

In the thread:

a lot of these photos lead to for-pay patterns that are also created by AI (or just stolen from actual artists) that won't even approximate what's in the photo (so scams)

gs17• 3 years ago

From the tweet:

a lot of these photos lead to for-pay patterns that are also created by AI (or just stolen from actual artists) that won't even approximate what's in the photo (so scams)

DougBTX• 3 years ago

The likes are probably not real then either.

reactordev• 3 years ago

Likes, praise, edification, dopamine hit for the reward center of the brain.

frostix• 3 years ago

The issue with generative AI techniques in general is how low the barrier to entry is. Various forms of information that used to be difficult or resource intensive to create have suddenly become approachable and even trivial in terms of resource investment to create.

Overall, in any sort of cost/benefit analysis, the cost is just so low now the benefits don’t have to be much of anything, if anything at all. Entertainment factor alone, boredom, or perhaps a passing curiosity to try something are enough to create and present false or misleading information and push it out to the public, creating noise needed to filter through. There are plenty of other far stronger motives that make the problem even worse.

Misinformation and disinformation were already becoming an increasingly large societal issue IMHO. That is only going to get worse with wide access to generative AI. We already have a high degree of erosion in social trust where we pretty much have to consider motives and driving forces behind every transactional relationship we have these days and we could at least use costs to help sort that mess out: why would someone bother investing the resources to do this? Does it cost a lot to present me with false information and if so, is there enough potential motive behind that to make this information more likely to be false or misleading?

The answer to this is increasingly yes. It’s now far more difficult to start from a position of distrust and move to a point of trust or likelihood of trust and I think we’re going to see that even more in all sorts of aspects of daily life. I now have to assume most pieces of information out there are targeting me and attempting to manipulate me in some way (more than before). I fear we’re moving to a model of free speech that will put more weight on “authoritative” sources more so than in recent past in many cases considering liabilities authorities have when presenting false, misleading, or inaccurate information. Liabilities that in many cases aren’t real liabilities just perceived liabilities, granting authoritative information sources far more credit than is due.

croes5• 3 years ago

What's the point of cheating in an online game or posting fake pictures on Instagram?

dkjaudyeqooe• 3 years ago

For people who don't understand what generative AI is and think it's real, the future is going to be strange and hallucinogenic.

songbird23• 3 years ago

i know of generative AI but dont know crocheting. I could not tell whether that image she posted was realistically doable by hand or AI

rblatz• 3 years ago

Look at the face, does it look like a photo with a yarn texture applied, or does it look made of yarn?

KineticLensman• 3 years ago

(my wife crochets).

Crochet figures generally look 'pixelated', where a pixel is a single stitch. The eyes on these figures don't look stitched at all. The hair looks like waves of material that also lack a stitched look (compared with the faces). The curves of the lower lip cut across stitches (going by the size of the stitches on the adjacent skin) and don't have that pixelated look.

[Edit] Also, I don't know what these crochet sites generally show, but in the pictures I've seen, it's usual to show the figure against a plain backdrop or on a table, not part of a scene with a detailed background (which in this case, also lacks stitches)

smt88• 3 years ago

This will be most people.

croes5• 3 years ago

Will be a feast for CTs.

Finally images and videos as "proof"

8organicbits4• 3 years ago

Is this really different from what we've always dealt with? We have ghost stories, drawings of monsters, fiction books, staged photos, etc. The history of fiction goes back as far as non-fiction.

pavel_lishin3• 3 years ago

Yes, but it used to take time to generate and spread.

Now, you can write a script that'll automate the entire process and generate more garbage in 3 hours than the entire online world could in 2002.

aaomidi• 3 years ago

The sheer scale of how fast AI can put out content is terrifying.

Think about all the work humanity has thought, written out, drawn, etc.

We can probably create an order of magnitude more than the entire history of humans within a few days/weeks.

This is truly revolutionary. This doesn’t mean what happened in the past wasn’t. Those had profound impacts on human life, again due to the scale of their impact.

But keep in mind those technologies being introduced was very painful for the people at the immediate time, and then slowly got easier and easier. We are the same people that AI is going to be extremely painful on.

3cats-in-a-coat• 3 years ago

Understanding it doesn’t help much. Our entire civilization has been on a path of generating fake everything for some decades now, and this is the final step

mgerullis• 3 years ago

With increasingly better AI, not even techies will be able to tell apart what’s real and what isn’t by just looking at it. You’re probably in for the ride.

skepticATX• 3 years ago

I think that tech companies will soon have to grapple with the fact that the majority of people currently have, or will have, predominantly negative opinions of generative AI.

We too often get lost in our tech bubble. Of course generative AI is useful for us. We also understand the nuances of the technology, so we can be more deliberative in its use. This is decidedly not what is happening with the typical person.

To put it bluntly, for all of the fanfare last year, how has generative AI had a positive impact on the average person? And how has it had a negative impact? I think that I can make a stronger case for negative impact (or no impact) than I can for positive impact, as it currently stands.

This is not inherent to generative AI; I can think of other technologies that followed a similar trajectory. But the difference is that generative AI leaders literally think that they are about to create a techno-utopia, and they may be sorely disappointed when others don't feel the same way.

Swizec2• 3 years ago

how has generative AI had a positive impact on the average person?

Generative AI, for the most part, is a technology in search of a problem. For now. The tech is in a “demos well, productizes poorly” stage. Once you try doing anything at scale and with proper evals of success, you see that real world performance is pretty poor still.

Yes yes we have lots of totally-not-cherry-picked papers where researchers achieved something fantastic. Then you look at the detail and it’s either “we ran this once because expensive” or “it achieved this great result almost 35% of the time so it’s state-of-the-art best-in-class”

create a techno-utopia, and they may be sorely disappointed when others don't feel the same way

Every techno-utopia I’ve ever seen in movies, books, etc has always been a secretly dystopia. Looks nice and polished on the surface, but achieves this result through aggressive oppression and disenfranchisement of dissenting voices.

Star Trek perhaps is the only one that didn’t follow that pattern. And even then you have groups outside the federation who are not super happy with how the federation does things.

phillipcarter3• 3 years ago

Just to offer a counter-point, we had measurable success with this tech solving a direct problem it's designed to solve for: https://www.honeycomb.io/blog/we-shipped-ai-product

My take is that the tech isn't in search of a problem, we're just early. It solves a lot of problems. The challenge is getting the right data pipelines in place to allow it to do that. The good news is that in many cases, it's a good thing to have that in the first place.

andsoitis• 3 years ago

Agree with everything you say, except perhaps the last sentence:

But the difference is that generative AI leaders literally think that they are about to create a techno-utopia, and they may be sorely disappointed when others don't feel the same way.

I would argue it is fairly common for people in tech to speak in messianic terms of the golden age our technologies will usher in. History is littered with this world view. Of course, there are also critics from within the tech sphere, but I would say the dominant mental model is one of naïve optimism that crosses over into arrogance.

bmer• 3 years ago

Is it the dominant mental model, or the dominant marketing model? After all, marketing must appeal to "techies" too, not just non-techies.

YeGoblynQueenne• 3 years ago

> We too often get lost in our tech bubble

A lot of the criticism of generative AI comes from techies though. I am one of them. I know there's plenty of others, on HN certainly, I think also on Twitter.

Personally, similar to your stronger case, I've now crossed a line where I find it more likely that generative AI will crash and burn and leave people shaking their heads at the misconceived notions of "making the world a better place" (the techno-utopias you say). We've had a year of crazy hype about GPT-4 and I still can't figure out what Gen AI is going to be good for, other than letting people rip off creatives, for fun or a very limited amount of profit; or for flooding academic venues with auto-generated work that will bog down the whole process of science.

danielbln2• 3 years ago

You're a techie and can't find legitimate usecases for GenAI? That honestly says more about you than it does GenAI. From coding to chat-with-your-data to all sorts of general data/text transformations, there is so much there.

Don't get me wrong, there are absolutely negative externalities, but if you can't see the useful cases as well then you're deliberately blinding yourself.

brucethemoose23• 3 years ago

There's a middle ground between utopia and scam-tech.

Generative AI is a powerful tool this very second, and is accelerating at an incredible rate. It seems pretty obvious that its going to automate tons of the white collar workforce. The hype will crash, no doubt, but its not going quietly into the night, nor is it zombifying like some other hyped tech.

Anyway, I view the whole AI vs anti AI debate as a distraction. Its coming, like it or not. The significant question is who will be in control... Will AI be local, distributed among small providers, or concentrated in a few pseudo monopolies?

phillipcarter3• 3 years ago

We've had a year of crazy hype about GPT-4 and I still can't figure out what Gen AI is going to be good for

Natural language interfaces to perform business-specific tasks, in concert with a UI to tweak things after generation, is the current killer use case in enterprise software, and it's exactly what's been built (and is being built). OpenAI's 1B ARR isn't just from people who pay for ChatGPT+.

GaggiX• 3 years ago

If we are talking about the real world I have never encountered a person with a negative opinion about generative AI, and I'm not just talking about tech people but people that mostly use ChatGPT for stuff like "some ideas for hiking trails in X".

CaptainFever• 3 years ago

Same! IRL everybody I know is at least positive to neutral on gen AI, with a few who don't like it. In school, it is very very widely used for obvious reasons, and I've seen a few friends use it for emotional support as well. The most opposition I've seen is always from online strangers.

raincole• 3 years ago

the difference is that generative AI leaders literally think that they are about to create a techno-utopia, and they may be sorely disappointed when others don't feel the same way.

No, the difference is generative AI guys believe that the content generated by generative AI will soon be literally (literal literally, not figurative literally) indistinguishable from those made by humans.

It's quite different from other techs. When people invented movie films, they (almost certainly) didn't think watching a movie will be literally the same experience as going to the opera.

Are they wrong? I don't know. But so far I incline to agree with them.

jstarfish2• 3 years ago

When people invented movie films, they (almost certainly) didn't think watching a movie will be literally the same experience as going to the opera.

People were actually diving out of their seats during the debut of Lumiere's "The Arrival of a Train." It was real enough to them...

xg15• 3 years ago

But the difference is that generative AI leaders literally think that they are about to create a techno-utopia

What honestly baffles me is that not even those are presenting a positive vision for AI.

On the contrary, the main PR angle of "generative AI" leaders right now is the "AI is a critical danger for humanity" doomsday cult. The one "positive" spin on this seems to come from the e/acc faction of the cult who frame it as some sort of natural evolutionary progression: "sure, AI will wipe out humanity, but that's actually a good thing".

But both the e/alt and e/acc legs of the cult seem to agree that a) AI is a threat to humanity, b) It's inevitable that AI will become more powerful than humanity and c) we nevertheless have to build it for some reason.

If the two publicly visible mindsets of its thought leaders are "cold-hearted Manchester capitalist who wants to replace even more workers with machines" and "tech-pyromanic madman who fears his soul to be eternally tortured by an AI unless he brings that very AI into existence" then it's not exactly a surprise if the tech is received negatively.

Baldbvrhunter3• 3 years ago

relax, it's temporary

"I crochet for clicks" vs "I crochet for enjoyment" has swung back to the latter.

So now hobbyists will have to build RL reptuation again, just like they did a few years ago.

The pattern scam sucks but, once again, reputation building will be important.

Hobbies will need to gatekeep. Post reels of work-in-progress etc.

The genie will not go back in the bottle - adapt or die I guess, sympathies for the dead.

ltbarcly34• 3 years ago

Agree, online spaces will adapt, and honestly how great was a community to begin with if as the only thing keeping it from becoming a hotbed of fake content is that the users didn't have the tools to fake it.

I'm sure very few people would agree when I say that nothing of value has been lost, but honestly I think this is more about people's illusions about the online communities they engage in becoming harder to maintain than those communities becoming worse somehow. I get the impression that people feel like their online spaces are being invaded by phonys using AI all of a sudden. If your community is so full of fakers and liars that it is dominated by fake content immediately when it becomes possible to generate that fake content, and that content isn't recognized by a critical mass of the existing users and laughed off the platform, you have been interacting with a lot of fakers and liars all along. That sucks, and maybe it was better psychologically to have that social outlet even if you only could only leverage it as a social outlet because you were being deceived, but there it is.

croes5• 3 years ago

How do adapt?

Everything can be faked.

jacquesm• 3 years ago

honestly how great was a community to begin with if as the only thing keeping it from becoming a hotbed of fake content is that the users didn't have the tools to fake it

It only takes a few users willing to post fake stuff to drown out a much larger number of real ones.

I'm sure very few people would agree when I say that nothing of value has been lost, but honestly I think this is more about people's illusions about the online communities they engage in becoming harder to maintain than those communities becoming worse somehow.

You could easily do the same thing with HN.

Note that real contributions and moderation don't scale whereas fake stuff does scale. So nothing is proof against an avalanche of fakes.

pardoned_turkey• 3 years ago

how great was a community to begin with if as the only thing keeping it from becoming a hotbed of fake content is that the users didn't have the tools to fake it.

Do online communities need to be great by some HN metric in order to be deemed worthy of survival?

Some knitters had fun sharing their work online. Now that's ruined by ML. If it was ruined by cryptocurrency mining instead, would we be posting similar "they had it coming" defenses of this destruction? There is a heavy bias on HN toward excusing the externalities of generative AI.

pavel_lishin3• 3 years ago

honestly how great was a community to begin with if as the only thing keeping it from becoming a hotbed of fake content is that the users didn't have the tools to fake it.

That's like saying, "how great was your house to begin with if the only thing keeping it from becoming smouldering embers was someone with a can of gasoline and matches?"

croes5• 3 years ago

Temporary like fake images with filters on Instagram?

adapt or die I guess it's die

safety1st• 3 years ago

There will be a certain percentage of the population, though, which fails to distinguish artifice from reality and essentially drives themselves insane by obsessively consuming lies. See the millions of women on Instagram posting photos that are very obviously doctored and filtered beyond all credibility and yet somehow inflicting depression and eating disorders upon each other anyway.

Uehreka• 3 years ago

adapt or die I guess, sympathies for the dead.

I dunno dog, you don’t sound super sympathetic.

I’m really tired of these takes where people look at a community they aren’t a part of that’s struggling with AI-based issues, hand-wave some non-solutions (usually involving some reference to “reputation”) and then shrug and go “and if you can’t solve these problems that society has never successfully solved then I guess you deserve to fail”.

Baldbvrhunter3• 3 years ago

I am sympathetic, and while I'm not part of crochet I am professional ceramacist and it's even worse for us. The generated pictures I've created of pots are indistinguishable from genuine ceramics - even to my eye.

There is no recourse, no genie back in the lamp.

Photos of strangers saying "I made this" and trusting that is over.

It's not about deserving, it's "deal with it"

ImAnAmateur• 3 years ago

There was this poor lady on a board who crocheted this damn homunculus when it was supposed to be this elaborate picture and she just got totally scammed! Waste of material and time and money!

I would love to see this. Both the instructions and the outcome. I have no clue what it would-- could!-- look like.

genman• 3 years ago

What happened with the tradition to add 1/N to tweets to indicate multi-post thread?

14• 3 years ago

Definitely a problem but will open up value to other contributors as well. Someone willing to make time lapse videos of their entire project or someone who creates a site where contributors are verified by different means like sending work to verified assessors who can say this is real will be allowed to sell verified patterns at a good price.

chrischen• 3 years ago

Found this gem yesterday while trying to research “k3s vs k0s” on Google: https://www.nops.io/blog/k0s-vs-k3s-vs-k8s/#:~:text=While%20....

Unsurprisingly it is on the blog of an AI company. The article contradicts itself multiple times, restating who is the developer of each of k0s, k3s, and k8s. Even if you are unfamiliar with these things a human being reading it can easily see it’s false information from the contradictions alone, which just goes to show this “nops.io” is just publishing AI blog spam without even having a final human proof read.

And the worst is that Google’s automatic algorithms are consuming it and thinking it is good content, making it the featured snippet.

dustincoates• 3 years ago

It could be AI, but content farm garbage has been around for a long time. I remember seeing one that was on how calculate total addressable market, and the whole post was "find out the total market, and determine what percentage you can address." It was back in 2018 or so, so someone had to put in the effort to write it.

drubio2• 3 years ago

Only now its more scalable to do content farm garbage with AI, it's cheaper SEO on steroids.

I'm optimistic it will one day be possible to sift through AI generated garbage, but it will take time just like it did with email/spam. And the most likely outcome will be through paid services, either paid content or paid filtering, just like email works best to this day.

I remember the early email days, early 2000's, pretty much anyone could setup their own email server (qmail/sendmail), there wasn't much spam to worry about and it required a lot of effort to make spam cost effective. Fast-forward today, even though you can still setup one, it requires a crazy amount of effort to ensure delivery in-and-out due to spam abuse, that, or pretty much paying a transactional fee, which is the easiest, so large providers don't flag email as spam.

justsomehnguy2• 3 years ago

But they have an audio version of the article!

drubio2• 3 years ago

The information super-trash-way, the data is feeding on itself.

We're heading back to the information dark ages. I don't know if I'm glad or sad, the pendulum is swinging the other way, where printed books or face-to-face learning, will come back in vogue to get vetted information.

zoklet-enjoyer• 3 years ago

I've been seeing this and fake furniture/rooms on Facebook ever since Midjourney got semi-realistic looking. It happens so often I got tired of calling it out. Some older people think it's real, but for the most part I think people know it's not real.

bakugo• 3 years ago

This is the inevitable fate of the entire internet if generative AI isn't banned.

Nevermark2• 3 years ago

a lot of these photos lead to for-pay patterns that are also created by AI (or just stolen from actual artists) that won't even approximate what's in the photo (so scams)

We need software tools where anyone can design their own ambitious but reliable designs.

But add to that machines that can create instances of those designs and I can see human crafting getting completely devalued.

We expected AI Spock. We got Robo Leonardo da Vinci.

harryf• 3 years ago

To prove we’re human, we need to behave in vulgar ways that generative AI is prevented from. https://youtube.com/shorts/uWgEofct9Rw?si=nE4lDqVYKa1a7XhR

In this case it would be crocheting a swear word on your design.

MintsJohn• 3 years ago

I'm more and more thinking a government issued digital identity (like https://privacybydesign.foundation/irma-en/) that can be used to proof you're human (and other details of you want) but that can't be tracked back to an individual, but again optionally, can be used to create (multiple) online persona's is the way forward. I used to think of these things as dystopian, but fake content by fake persons is a bigger issue. Of course real persons could create persona's for such a bot, but a (personal and/or community based) blacklist mechanism based on the root account (the real human that created the persona) would go a long way.

emsign• 3 years ago

Yes! I've noticed some decay in content quality at different places recently. Search results have become especially bad, there's so many content farms with AI generated nonsense now. And what's funny is that helpers like Bing Chat are searching the internet to give answers.

There's an upside though, this makes me use the internet and my smartphone less and less. And I'm sticking to very specific sources now.

Time for me to stock up on some HDDs to create a personal archive of data that hasn't been touched by AI yet...

yetanother12345• 3 years ago

Amigurumi (Japanese: 編みぐるみ, lit. "crocheted or knitted stuffed toy") is the Japanese art of knitting or crocheting small, stuffed yarn creatures.

Had to look that up

https://en.wikipedia.org/wiki/Amigurumi

CrzyLngPwd• 3 years ago

It's going to get far worse yet, and there isn't anything that can be done about it.

Hopefully, when we are sick of this fluff, it will encourage people to step away from their devices and to experience the real world more.

Generative AI flooding online crocheting spaces with unrealistic amigurumi pics

Related Stories