Generative AI flooding online crocheting spaces with unrealistic amigurumi pics
twitter.com
twitter.com
So your insistence on 'privacy' yet 'only vetted by Mozilla' is misguided at best.
Re. Mozilla vetted extensions, I cited security, but would love to learn something new about privacy if you'd like to share more.
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.
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.
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.
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...
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)
For the hair, I guess it didn’t really stand out, I’d believe it was possible to get pre-made doll wigs or something like that. Thinking about it, though, it does seem a bit odd that it would look like bundled yarn up top and then more like hair in the back.
For the hands, it is kind of funny, I’m not sure if it is an artifact of the typical “AI is bad at hands” thing or “intentional” (to the extent that the output of a ML model can be called intentional), but the hands do appear to have some mistakes.
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.
As an industry and artifact, AI is rapidly filling the gap between what our brains “measure” (novel, obnoxious, sexy, …) vs. intended outcomes (learning, threat mitigation, healthy mates, …)
Culture leveraged that mismatch.
But that gap has now been colonized by a new species class: the uber adaptive, fast generation, artificially intelligent attentionvore.
The farming of people has just begun. The next Facebook is coming. :(
EDIT: Also, you don’t actually have to write that script now…
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.
The industrial revolution after all was unprecedented in the history of human civilizations until it happened. Nobody had a roadmap on how it will all play out.
Finally images and videos as "proof"
I would love to see this. Both the instructions and the outcome. I have no clue what it would-- could!-- look like.
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.
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.
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?
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+.
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.
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.
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.
I wasn't trying to say it's useless, I was getting at this sentence from your 2nd paragraph: "[did it work?] The short answer is... mostly."
That "mostly" part is the cold reality many products have run into this past year. It's been the biggest problem in my experiments at least. The current tech requires a lot of babysitting.
> 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.
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.
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...
Actually, they probably weren't. This story appears to be an urban legend, and there's no evidence that it ever actually happened[0].
[0]https://www.atlasobscura.com/articles/did-a-silent-film-abou...
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.
Had to look that up
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.
In this case it would be crocheting a swear word on your design.
> 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)
> 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)
And maybe internet points.
Unfortunately, there are a lot of people who don't have any pride in their work.
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.
...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.
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.
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.
"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.
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”.
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"
>adapt or die I guess it's die
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.
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.
Lets say you go to the local comedy club and do your act. You aren't a professional comedian, but you want to do comedy for fun, and it's a social experience where you get to interact with people. You work from home and social outlets are very important!
You do this for years, getting feedback from other comedians and enjoying yourself. Later you decide to buckle down and watch the great comedians at work. You watch the top performances by the top comedians, and to your horror you notice something disturbing. The other 'comedians' at your social club are just reading lines they copied down from famous comedy acts they watched on Youtube, verbatim. In fact, out of hundreds of other amateur comedians you had been having drinks with, and being proud when they praised your act or gave feedback, maybe one or two of them were actually writing their own material. They are all faking it to get social credit in this club.
At what point was the club ruined, all along or when you found out?
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.
Everything can be faked.
That's not adapting but retreating.
Everyone will have a chance to be the big fish in a small pond again, rather than being a drop in a global ocean. Everyone's self-esteem will improve as a result. Nobody will feel inadequate compared to everyone else on the planet. Nothing is more demotivating than seeing leaderboards dominated by random Japanese kids that you'll never have a chance of ever beating.
Yeah, there may be a knitting goddess in Ohio you can't reach out to but you'll foster a deeper relationship with the hobbyist down the street, which would arguably benefit you more anyway.
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?"
It is more like you hire someone to repair a pipe and when they open the wall they find mold everywhere. You had been happy living in the house for the last 15 years, but now you know about the mold and you feel grossed out being in the house.
I think these forums are going to discover that a different, more stringent kind of moderation is needed now that doing this kind of shit is cheap. I guess maybe it's more like you've been living your house happily for 15 years, but with the change in climate, suddenly your house is full of a kind of rodent that hasn't been able to survive in your region, and you're just shoveling out buckets of rodent poop every day, and trying to figure out what kind of traps to set out to stop them from coming in.
(Except the rodents aren't just appearing naturally; your neighbor started a rodent breeding farm, and is releasing them into your house so they don't have to pay for extra food or space, and is selling them for a profit while your house gets slowly filled up with rat shit.)