OpenAI’s new AI image generator is potent and bound to provoke
- Thread starter JournalBot
- Start date
When asked to present evidence to back up your prediction that AI will kill all jobs - you bragged about your "broader perspective, good education, curiosity, imagination and common sense. ".
If that isn't hubris...
Upvote
-11
(0
/
-11)
It won't generate an image of Trump or Musk, at least based on a short trial. Anyone have a prompt that works?
Upvote
0
(0
/
0)
I don't know what you mean. People choose to deliver pizza and your groceries for a pittance plus tips rather than take skilled and professional jobs. So much more rewarding to be terrorised by statistical judgements of performance from an automated manager and eat shit from the general public.
Upvote
-9
(0
/
-9)
Then again, maybe some people know a thing or two because they saw a thing or two? Like certain precedents where decisions affecting lives of actual people were taken based on tenuous or ouright improbable business cases? Or where people were let go because some unimaginative high management types considered cheap replacements to be "good enough", customer satisfaction/retention be damned?
How many serious analyses of the Ai impact on employment have you seen so far? How many assessments of demand/consumption decreases? Anything about potential inequality and/or poverty implications? Where would the return on the mind-boggling capital investment come from? How do the AI companies propose to actually earn the money to keep their datacenters running?
So forgive me when I'm less than impressed by shiny new ghiblification being the future.
Upvote
9
(10
/
-1)
Isn't the whole point of all of our technological progress to make things easier so people have to work less at things? Provide people with more capability? Let's say I want a graphic to make my corporate newsletter more interesting. Now, I can get that fairly easily. That's a good thing! Yeah, I don't care to improve my artistic talents, but I have other ways to personally grow in areas that I'm actually interested in. I understand the concerns regarding misinformation and other malevolent uses, but I'm just not tracking on this. Just because AI image creation exists doesn't mean you have to stop any sort manual art hobby, etc.
Upvote
-1
(3
/
-4)
No, it's experience. Actual policy and decision makers are rarely altruistic and they will rarely work for the good of the people that voted them in. If given the option to act in their own interest, they will do so. And I don't even mean the current situation in the US; I'm not even American and maybe - just maybe - I have a slightly more grounded perspective that makes me wary of this handwaving about the future.
Because get this: I'm not afraid of what is currently called "AI". I'm afraid of the people that will blindly trust in it and will not even have the slightest idea when their magic black box fails them.
(By the way, I don't think you presented any evidence that AI won't kill jobs. But hey, hubris and all.)
Upvote
1
(3
/
-2)
Yeah, part of the problem graphic artists, commercial writers, catalogue models, etc., are currently facing is that everyone unaffected was fine with automating and offshoring blue collar work, then the clerical work and call centre work to which many of those people migrated. Oh, there was plenty of hand wringing from some, but no follow through.
Then people who definitely knew better acted all surprised that not needing those workers meant all the supporting white collar workers weren't needed either. Acted all surprised that there was less wealth and consumption in the communities, leading to services, and those jobs, disappearing too.
With Amazon eating retail, what's left for some people? Logistics and last mile delivery? For now. Graph goes up for Wall street at the expense of graph going down for Main street. Those gains at the top don't trickle back down into communities either, they are invested globally or put into the latest tech bubble. We can't live by selling each other sandwiches and giving each other rides people, even though that may make GDP go up - the measures of national economic success are no longer moored to the experiences of the people.
Upvote
5
(6
/
-1)
To deal with the tsunami of fake images coming from this (and other) models, prehaps we can start going back to tangible physical originals to prove image authenticity (i.e. a film negative or a slide). Very hard to argue with an intact physical original.
Upvote
-3
(0
/
-3)
I think the original query was something like, do that meme Ghibli style. Did you now elabotrate the query to specify that they guy needs to look lecherously/whistle? Because the idea is that the AI will figure it all by itself.
My point was that that the AI did not know what to do with the expression in the meme photo, and does not understand how essential it is, so it picked a random one from what was available. These two here are not as clueless, but they don't look like Miyazaki anymore.
Likewise, the girlfriend's eyes are pointing straight (though her head is turned), because that's a characteristic "angry girl" expression in Miyazaki. They do not shoot daggers at the guy, as in the meme photo, because the AI has not figured that part out.
Upvote
2
(4
/
-2)
Yes, it's bollocks. Many people will play around with image generation occasionally for fun or hobbies. Work presentations will get even more annoying, and small businesses who want it will be able to afford custom artwork (although I'm sure paying a final year art student or reaching out on Deviant Art wouldn't break the bank), but the real impact will be felt by commercial artists losing their jobs as capitalism does its thing.
Upvote
6
(6
/
0)
So rather than throwing out the technology, shouldn’t we be evolving our society with it? That’s what we should be pushing for! Technology is supposed to increase quality of life for humanity. The bad thing here is not the tech, but society not adapting to truly benefit from it! We need progressive policies damn it!!
Upvote
1
(3
/
-2)
Policy makers are always selfish. And you're right to be wary.
I have no quarrel with someone who is afraid of AI potentially taking jobs and the kind of future that means. That is a real worry and I share it.
But let's not pretend to know what's going to happen here...
I don't have a problem with your position, I have a problem with the certainty with which you've predicted the future.
Sure. Again, I share your fears and worries...
You still don't get it. I'm humble enough to say I don't know what will happen. You could be right or wrong.
History is definitely not on your side. No new technology has led to mass unemployment in history.
Now, this technology is nothing like anything before so you could be right but don't act like you have some sort of clairvoyance in the face of technology humanity has never seen before.
We are off the edge of the map here... Here there be dragons...
Upvote
-6
(2
/
-8)
He's a salesman, a very good one. He spent all this time bullshitting to make money.
Upvote
5
(5
/
0)
Is it?
How is that borne out do you think? Do you feel like people work less now than they did 20 years ago? Do you think people have more free, time, a better standard of living?
I don't believe that is in fact the point. We're good at extracting shareholder value. I don't think we're particularly good at making things easier for people.
And you cannot divorce the basics of capitalism from this conversation.
Why is it a good thing though?
If you were to break down what made it good what would your basis be?
I mean, I could give you an example: you didn't have to pay anyone to do it for you. Now that's part of the cycle of things to an extent, but we're seeing an unprecedented shift here, and people are quite rightfully nervous about what it means. Because again, capitalism. This isn't in a vacuum.
Let's consider the ethics though. Where did that image come from? Because all these dumb anime memes? They're all ripping off a studio. It's pretty shitty. If a human artist was stealing this blatantly they'd be shunned. But OpenAI has it baked into their business model to steal. I find it despicable personally.
But yay, you put some art in your corporate newsletter.
You don't care to grow. You just want the labor for free instead. Do you see how that might feel problematic to people?
Upvote
10
(12
/
-2)
Hate to break it to you, but cash register jobs are gone too. Either pay via app or scan the groceries yourself with one person watching 5 to 30 registers.
Upvote
6
(6
/
0)
Even if you are uniquely skilled. Before mass production, mass markets, and IP rights, those writers, composers, and artists whose work you love either had another income source or were paid not much more than a basic living by patrons
Upvote
5
(5
/
0)
You are still attacking the technology instead to attacking the lack of change in society to benefit from the technology. That is my point. It’s not the technology that is bad, it’s the people that are holding us back from actually benefiting. Also, I didn’t say that I don’t care to grow. I can still grow despite having new capabilities. I don’t care to grow my artistic talent, but there are so many other areas one can focus.
Upvote
-6
(2
/
-8)
Okay, so I have two problems with your response to me.
First off, the technology is bad. The ethics of it stealing from people suck. OpenAI doesn't try and hide this.
OpenAI declares AI race “over” if training on copyrighted works isn’t fair use
It's not fair use. It's theft. They're trying to play by different rules than everyone else.
Secondly, I'm living in the real world. Get back to me when we're in a post capitalist utopia and we can discuss whether or not I feel differently. In the meantime we live in this society, and I'm going to react within the context of that.
Upvote
9
(10
/
-1)
Again, same question, the probability of WHAT? And how are those probabilities calculated?
If I paint a painting, I can describe what I'm doing as a series of actions that I estimate as having high probabilities of achieving my desired outcome.
Does that mean I've never "learned" anything? Does it mean I don't "understand" what I'm doing? Does it mean that what I'm doing is not artistic, because probabilities are involved?
Again, throwing the word "probability" into the discussion by itself is literally meaningless.
Upvote
-8
(0
/
-8)
A good friend is in the same position, other than perhaps having longer to retirement and desperately weighing up a career change. She is somewhat lucky in being skilled enough to split her time between complex texts and correcting the work of juniors. One of her laments is that newer entrants to the field seem to have relied on technology for their studies and often don't grasp the basics of the language or the profession.
Upvote
8
(8
/
0)
Thank you for reminding me of a variation that a friend used in their dot-signature, a nearly depressing number of years ago:
"A foolish mind is the consistency of little hobgoblins." [from Tamsin's .sig, some time in the early 90s]
Upvote
2
(2
/
0)
The other part of Distracted boyfriend which appears to have been missed by both the original AI art and the ensuing discussion is that the lady in the red dress is completely oblivious to the other people in the picture, whereas all the AI has her looking back at them.
Your fuzzy and artsy question is actually the key one from an AI standpoint as well. The continuing issue with these image generators is that they’re drawing 3d functional objects having only trained on 2d rendering of objects with no knowledge of their function.
The axe’s handle doesn’t match where the head is drawn, because the AI doesn’t know enough about tools or drawing in 3d to get the line right. The sword has a weird lopsided hilt because it doesn’t understand that a sword of that type will be symmetrical in order to be usable. The TV has eight buttons in one corner with no suggestion of their function, and is showing a blue screen for some reason even though TVs don’t do that.
There are some known triggers for for this issue, and it would be interesting to see if they still work. Getting a visual AI to draw a glass of wine full to the brim, or a clock at 3:30, because it hasn’t trained on those and it has no mechanism for understanding how they would be different. Getting consistent shadows (like in the dog picture) or getting houses that look like functional houses are other difficult ones. They’ve mostly fixed the hands, via a massive amount of training on hands, but they’ve got so many more concepts to do that on to fix all these issues.
Last edited:
Upvote
6
(6
/
0)
Again, you are conflating the tech with the societal stuff around it. The tech is fine, your issue is with what was put into he models and how.
Who is everyone else? China?
Sorry, we're just born into that transitional phase (society/governance catching up with tech.), hopefully progressing towards that utopia. Generation Beta will grow up with AI right by their side. They'll embrace it like GenZ seems to embrace the Tikitiy Tok. They won't know a different pre-AI life. Yes, living through the societal transition to utopia will never be as good as the utopia, but we should be striving for future generations to have it better than us.
Upvote
-11
(1
/
-12)
Except in the current analogy we're not the buggy drivers....we're the horses.
Upvote
7
(7
/
0)
Where do you live that you see us currently striving for Utopia?
Upvote
5
(5
/
0)
As long as we live in society I'm utterly uninterested in hand waving that pretends we don't. There's no point in discussing how a bullet would to the head would be fine if you kept your brain in your ass instead.
Me. You. Who said anything about China?
Mate, if you think my kids are growing up in a better world than I did you are definitely living somewhere else.
Anyways, moving on now.
Upvote
10
(10
/
0)
Not to start a rehash of comments on copyright, but the US legislation put it well. The spark of human [vision and] creativity is the difference.
Cross out human if you like, but where is the guiding creative mind?* It isn't in the prompts, unless you are saying it was the patrons who created the ceiling of the Sistine Chapel or the Mona Lisa.
I think LLMs are a good tool for a number of purposes (inspiration, mock ups, early drafts in creative fields), but you can't divorce a tool from the uses to which it is put, the societal context and impact.
*For those who think LLMs are aware. The only consciousness we know of is based in physical brain matter. People are putting distilled versions of last years biggest LLM models on Macbooks, where is there the room for consciousness to exist in those downloaded statistical charts and on that hardware? There is no evidence of consciousness in the outputs, just lots of anthropomorthising of limited technology and the field's use of words that seems designed to lead people to assume consciousness.
Upvote
4
(4
/
0)
I agree.
I'm completely open and curious to using AI tech to enhance human creativity. When you're using it to replace it I check out. "Prompt engineer" sounds like a title some Hollywood prick made up to get into the credits of something they didn't do any work on.
Upvote
6
(6
/
0)
5 years ago this might have been an accurate description of what was happening.
Google Translate had switched to a deep learning model, vs. Bayesian inference.
Bayesian inference was basically crap. It was impressive when it translated anything correctly at all.
The Google Translate deep learning model was remarkably good compared to Bayesian inference. That's when human translators started to get worried. It usually produced accurate translations, but it wasn't hard to find examples of poor translations because context couldn't be communicated to the system, or just because it sounded robotic or artificial. It was easy to make the case that human translation was better.
Now, it's pretty easy to argue that the best LLMs do a better job of translation than maybe all but the very best humans.
I know people who do professional translation and they are now relying heavily on ChatGPT to suggest better translations than what they can come up with, and to explain linguistic subtleties to them that they aren't aware of, even if they speak both languages as mother tongues. The job that ChatGPT does can hardly be described as "half-assed."
Upvote
2
(2
/
0)
I would love to see your evidence that cashiers and delivery drivers do so as their first career choice, the delivery drivers I've known did so because it was their only option other than starvation
Upvote
4
(4
/
0)
"Sometimes I had to earn extra money," Putin said. "I mean, earn extra money by car, as a private driver. It's unpleasant to talk about to be honest, but unfortunately that was the case."
Upvote
2
(2
/
0)
err, yes, that's the point I thought I illustrated. Unfortunately, as per my other post, you can't divorce a tool from the uses to which it is / will be put and the societal impact. The impact on the world as we know it, not a notional world assuming changed variables.
Not sure progressive policies gives the meaning I intend. I like democratically regulated capitalism, with a social safety net / springboard and a recognition that society is not all about profit and loss. I think the economy should be run to serve the community, not the community run to serve the economy.
Upvote
0
(0
/
0)
Upvote
-5
(0
/
-5)
Europe and Canada are at least somewhat trying. Not every country is as regressive as the USA.
Upvote
1
(2
/
-1)
I really, really do. Because if you ask my opinion wearing my real life artist hat [nominally], I’ll direct you at the one with two eyes, ofc, because [with some specific exceptions] people aren’t cyclops.
The other only functions because [again excepting exceptions] human brains learn to implement object persistence. And facial recognition is so important to us as a social species we master it earlier than continence. Thus at our age, we know so totally that a human face has two eyes, if see only one we fill in its partner implicitly. Don’t even think about it. Completely automatic.
My point being: it really doesn’t matter if that AI genny was wrong, from a single perspective. ALL 2D representations are absolute cheats. They are ALL dead wrong. BOTH images I showed you are your “abstract portraits,” both abstractions employing one set of lies or another to fool us observers.
“Fails to conform to a set of geometric rules for 2D projection” is more than reductive, it’s missing that point completely. Even once its light has entered your eyeballs, what you “see” is still only synthesis. A fabrication derived by complex wetware running “Human v1.1” simulation. Whatever works is whatever communicates. There’s myriad available perspectives an artist may play with, but the only interpretation that matters is its viewer’s. If Art doesn’t throw a wrench in this damp machinery just to discover what happens, what else is it good for? Comfort eating?
…
I recommend reading this. It’ll give you heebie-jeebies realizing how much of your “Real” = unaware, unquestioned, barely tested assumptions. Good book.
TL;DR: Don’t look at the “art” to get its formal measure; look at the viewer’s response to it.
STL;SDR: Technical precision’s for dick-swinging rights; neither true nor very interesting.
--
“We all see what we want to see. Coffey looks and he sees Russians. He sees hate and fear. You have to look with better eyes than that.”—“Lindsey Brigman” (Mary Elizabeth Mastrantonio), The Abyss
Upvote
-6
(0
/
-6)
Thanks for posting this.
People worry about the use of AI for disinformation, when the problem is not the quality of the disinformation but that people will spread anything without checking.
Upvote
0
(2
/
-2)
Respectfully, why can't you learn to be a story teller with AI? I saw a 3.5 minute long short movie today that was built by one man in ~2 weeks using a mix of models to get from ideas to sketches to image to video, playing all the roles himself with controlnets and voice shifters and music and whatnot. He didn't have actors or a set or a cameraman or lighting or sound crew or costumes or props or CGI or anything, but he had a good story, quite a bit of talent and a whole lot of GPU hours. I thought it was impressive.
Sure, you might say the AI had quite a bit of creative control in the final result, but it was his story not the computers and the alternative would be trying to get a budget and friends together to LARP it with all the constraints and them potentially wanting to put their own spin on the story. Because the AI might be dumb but it also means it has no ego, if you don't like what it did just change it until you're happy with it. It's not like the real world has infinite time or patience with you wanting a 13th take to get that timing or delivery or expression perfect.
Upvote
-2
(1
/
-3)