In addition to the art theft and creeping laziness (to be followed by the inevitable acceptance of bad art as Normal), we have the energy use issue. This dreck uses a crapton of power. I fear it’s a losing battle, though.
Here’s my concern about AI, aside from the theft. Right now it’s appropriating from art created by humans. As it becomes more popular and ubiquitous, the sources available to steal from will increasingly be … AI. We will likely descend into a nadir of artistic homogeneity; humans will notice the problem but the monetary advantages of automation will weigh heavy against efforts to inject freshness.
Disagree to the extent that while there will always be a need for ordinary, workaday entertainment products which AI can crank out under licensing arrangements with its creators, I suggest this will, in fact, make the need and competition for, new ideas --- that "freshness" you mention --- all the more necessary and valuable. A working creative, presumably freed from having to grind out yet another potboiler to make a living can focus on those new ideas or approaches. Now, not all of these new ideas or approaches may come on a standalone basis and may be incorporated into AI datasets, but they will become unique selling propositions, especially as the business becomes more segmented or specialized.
Perhaps. It’s my concern, not my ironclad infallible prediction. But I do believe shrinking the market for human illustrators will make it harder, not easier, for the ones that remain.
The ones who remain will be in two general and not entirely exclusive categories: those living off so-called residuals from previous work being regularly regurgitated in AI datasets and/or providing the same ol' same ol' to those datasets and those who sit at the top of the creative chain generating those new ideas. Will some artists be replaced? Yes, but it would be less a matter of the technology than their place in the commercial pecking order and their ability to adapt. In other words, ability/salability and ability to adapt.
I saw a comment on some social media platform that said something to effect that AI has inadvertently provided evidence of the human soul by showing us what art looks like without a soul.
Note that the “marriage on the moon” AI rendering clearly shows the moon in the sky, an idiotic error that no human artist would have made. If the couple is standing on the moon, the object in the sky would be the earth.
From the Web: When viewed from the Moon, Earth would appear primarily as a blue and white sphere, with the white representing clouds, while the blue represents the oceans, making it look predominantly "white" due to the combined reflection of light from both the clouds and the water surface; essentially, it's a mix of colors that our eyes perceive as a whitish hue.-- This must be right b/c I am quoting AI.
I can't get beyond the fact that they're not wearing space suits. How are they breathing? How are they surviving the extreme temperatures and radiation?
Fact is, down through the ages, human fine and commercial artists have made "idiotic errors" either under the cover of artistic license or as outright blunders.
Humans make soulless art, too. At least the AI has a good excuse. Thomas Kinkade...eh... not so much.
I can only wonder if a one-in-a-hundred million AI could be as exceptional as a Picasso or Van Gogh. I suspect something will have had to have gone very wrong with the AI for that to happen.
I don't think that is the same thing. At least it isn't what I meant.
For that theory to be parallel to my imagining, it would need to posit that eventually a monkey actually as talented at writing as William Shakespeare would occur, and not merely suggest that random chance will eventually lead to the words on the pages being the same as Shakespeare's words. I don't know that I grasp the legitimate academic understanding of IMT as I've really only previously considered it as comedy.
I guess my implication was that the humans who make the art which touches our souls are not merely atypical, but kind of broken. And I wonder if the AI would need to be broken to achieve anything of a similar exceptionalism.
The IMT is of course a thought experiment, and considering the probability, although not zero, is so vanishingly low it would be realistically impossible. For those still clinging to the possibility, I would direct them to a smaller scale, but telling, experiment, conducted by students and faculty from the University of Plymouth at a zoo in Devon, England. When provided with access to a computer keyboard, the macaques involved managed to generate only random letters and primarily "S" --- in addition to using the keyboard as a toilet. Whether the recurring "S" was shorthand for their view of the experiment and/or the technology --- also physically expressed by taking dumps on the device --- will remain forever unrevealed.
I have that copy of the Obama New Yorker tucked away in a drawer, along with the crying Beefeater from when Princess Diana died, and the great black-on-black skyline from the Blackout.
I tried making a AI image once. I needed to illustrate a camel going through the eye of a needle (for Jesus's striking image). And AI couldn't do it. The best I got was a camel knitting a sweater. Also, your AI image of the wedding on the moon, note that AI put THE MOON in the background, not the planet earth! Oops!
My one attempt to create an AI image was an attempt to get a rendering of a person stealing eggs from a bird's nest. No matter what prompt I tried, I could not get the AI to understand that I didn't want Easter eggs. That is, I wanted normal bird's eggs, all the same color. The more I tried, the worse it got, and I finally gave up.
The only issues I see with AI-generated commercial art and entertainment are practical ones: the unauthorized use of work or likeness, copyright infringement per se and compensation. It's not as if the folks who train AI systems blindly select training data, they know (or should know) in detail what make up their datasets and should therefore know exactly what the data sources are and their specific content. Certainly a form of licensing similar to one or more variations of it used in the music industry should be, and no doubt is being, considered. This, of course, can get contentious because of ownership issues and licensing arrangements which may already be in place. Going forward, these matters will very likely be addressed specifically in usage agreements and similar contracts. As for the creative side, I believe AI can be a useful tool for artists and in fact, can provide opportunities to more easily experiment with, and expand the expression of, the "ideas," which form the creative underpinning of art. If it can serve as a labor saving device to free up a creative to focus on creating, all the better, since there is almost always a good deal of trial and error involved in the process. I can also see a future where consumers are the creators, or perhaps more precisely, the "constructors," of many creative consumables based on licensed components or datasets created from the work of professionals. Want a film noir movie directed by Billy Wilder with a plot by Gene Weingarten featuring a young Al Pacino in the lead role and set in Brooklyn (okay sorry, the Bronx). You got it. How about sequels to "Wuthering Heights" or "War and Peace" by Brontë and Tolstoy themselves? Okay, forget "War and Peace." As the Wise Man of the Bronx points out we're talking about "ideas" --- whether expressed as characters, plots, images, brush strokes or a specific arrangement of notes, performed in a certain way. What it will boil down to is whether AI can get those "ideas" straight enough, at least to the point where you are able to suspend disbelief sufficiently to believe you are reading an original or seeing a professional presentation. Or whether you care. It's not a matter of forensic analysis afterall, it's for enjoyment. Beethoven's "10th" is a case in point. It sounds like Beethoven, but to the trained ear of the aficionado or musician it is missing that "special sauce." Yet it is close enough to simply enjoy in the absence of any more original works to come for a lover of Beethoven.
Looks like the WaPo brain drain continues with political correspondents Ashley Parker and Michael Scherer reportedly leaving to join The Atlantic. Josh Dawsey is also said to be on his way out or at least seriously considering an exit.
I have seen a number of works done in the style of another artist. (Jack Davis of Mad Magazine fame seems particularly popular right now.) Is this any more theft than using AI?
The “borrower” in these cases like Mad usually sign their names. Ai seems not to ever do so. (I’m now using the small “i” so I stop reading that someone named Albert is ruining our lives.)
The genie is out of the bottle. We can wring our hands but AI is here to stay. Legislative guardrails will not have much effect.
As we’ve already seen in political campaigns, the use of AI, especially photos, will result in ever more hard-to-disprove disinformation/lies. How will it affect the admissibility of photos as evidence in trials?
What would an AI stairway drawing in the style of MC Escher look like?
In addition to the art theft and creeping laziness (to be followed by the inevitable acceptance of bad art as Normal), we have the energy use issue. This dreck uses a crapton of power. I fear it’s a losing battle, though.
I think the word "crapton" should be the word of the year -- probably next year, the way things are going.
Here’s my concern about AI, aside from the theft. Right now it’s appropriating from art created by humans. As it becomes more popular and ubiquitous, the sources available to steal from will increasingly be … AI. We will likely descend into a nadir of artistic homogeneity; humans will notice the problem but the monetary advantages of automation will weigh heavy against efforts to inject freshness.
Disagree to the extent that while there will always be a need for ordinary, workaday entertainment products which AI can crank out under licensing arrangements with its creators, I suggest this will, in fact, make the need and competition for, new ideas --- that "freshness" you mention --- all the more necessary and valuable. A working creative, presumably freed from having to grind out yet another potboiler to make a living can focus on those new ideas or approaches. Now, not all of these new ideas or approaches may come on a standalone basis and may be incorporated into AI datasets, but they will become unique selling propositions, especially as the business becomes more segmented or specialized.
Perhaps. It’s my concern, not my ironclad infallible prediction. But I do believe shrinking the market for human illustrators will make it harder, not easier, for the ones that remain.
The ones who remain will be in two general and not entirely exclusive categories: those living off so-called residuals from previous work being regularly regurgitated in AI datasets and/or providing the same ol' same ol' to those datasets and those who sit at the top of the creative chain generating those new ideas. Will some artists be replaced? Yes, but it would be less a matter of the technology than their place in the commercial pecking order and their ability to adapt. In other words, ability/salability and ability to adapt.
I always think of this story in relation to the value of creative art.
https://d-a-v-e.org/wp/wp-content/uploads/2020/09/The_Great_Automatic_Grammatizator.pdf
I saw a comment on some social media platform that said something to effect that AI has inadvertently provided evidence of the human soul by showing us what art looks like without a soul.
Note that the “marriage on the moon” AI rendering clearly shows the moon in the sky, an idiotic error that no human artist would have made. If the couple is standing on the moon, the object in the sky would be the earth.
they are standing in a moonscape.
It is ambiguous, like all good art!
The moon in the sky is completely unambiguous. And completely inconsistent with the idea that they are standing on a moonscape.
From the Web: When viewed from the Moon, Earth would appear primarily as a blue and white sphere, with the white representing clouds, while the blue represents the oceans, making it look predominantly "white" due to the combined reflection of light from both the clouds and the water surface; essentially, it's a mix of colors that our eyes perceive as a whitish hue.-- This must be right b/c I am quoting AI.
Touché
I can't get beyond the fact that they're not wearing space suits. How are they breathing? How are they surviving the extreme temperatures and radiation?
Fact is, down through the ages, human fine and commercial artists have made "idiotic errors" either under the cover of artistic license or as outright blunders.
Also, the groom looks more like Tom Brady than Travis Kelce.
Humans make soulless art, too. At least the AI has a good excuse. Thomas Kinkade...eh... not so much.
I can only wonder if a one-in-a-hundred million AI could be as exceptional as a Picasso or Van Gogh. I suspect something will have had to have gone very wrong with the AI for that to happen.
Yeah, the Infinite Monkey Theorem. Will that devolve into the infinite AI theorem?
I don't think that is the same thing. At least it isn't what I meant.
For that theory to be parallel to my imagining, it would need to posit that eventually a monkey actually as talented at writing as William Shakespeare would occur, and not merely suggest that random chance will eventually lead to the words on the pages being the same as Shakespeare's words. I don't know that I grasp the legitimate academic understanding of IMT as I've really only previously considered it as comedy.
I guess my implication was that the humans who make the art which touches our souls are not merely atypical, but kind of broken. And I wonder if the AI would need to be broken to achieve anything of a similar exceptionalism.
The IMT is of course a thought experiment, and considering the probability, although not zero, is so vanishingly low it would be realistically impossible. For those still clinging to the possibility, I would direct them to a smaller scale, but telling, experiment, conducted by students and faculty from the University of Plymouth at a zoo in Devon, England. When provided with access to a computer keyboard, the macaques involved managed to generate only random letters and primarily "S" --- in addition to using the keyboard as a toilet. Whether the recurring "S" was shorthand for their view of the experiment and/or the technology --- also physically expressed by taking dumps on the device --- will remain forever unrevealed.
I have that copy of the Obama New Yorker tucked away in a drawer, along with the crying Beefeater from when Princess Diana died, and the great black-on-black skyline from the Blackout.
Those plus the 9/11 one
I tried making a AI image once. I needed to illustrate a camel going through the eye of a needle (for Jesus's striking image). And AI couldn't do it. The best I got was a camel knitting a sweater. Also, your AI image of the wedding on the moon, note that AI put THE MOON in the background, not the planet earth! Oops!
My one attempt to create an AI image was an attempt to get a rendering of a person stealing eggs from a bird's nest. No matter what prompt I tried, I could not get the AI to understand that I didn't want Easter eggs. That is, I wanted normal bird's eggs, all the same color. The more I tried, the worse it got, and I finally gave up.
Camel knitting a sweater is not bad, though.
The only issues I see with AI-generated commercial art and entertainment are practical ones: the unauthorized use of work or likeness, copyright infringement per se and compensation. It's not as if the folks who train AI systems blindly select training data, they know (or should know) in detail what make up their datasets and should therefore know exactly what the data sources are and their specific content. Certainly a form of licensing similar to one or more variations of it used in the music industry should be, and no doubt is being, considered. This, of course, can get contentious because of ownership issues and licensing arrangements which may already be in place. Going forward, these matters will very likely be addressed specifically in usage agreements and similar contracts. As for the creative side, I believe AI can be a useful tool for artists and in fact, can provide opportunities to more easily experiment with, and expand the expression of, the "ideas," which form the creative underpinning of art. If it can serve as a labor saving device to free up a creative to focus on creating, all the better, since there is almost always a good deal of trial and error involved in the process. I can also see a future where consumers are the creators, or perhaps more precisely, the "constructors," of many creative consumables based on licensed components or datasets created from the work of professionals. Want a film noir movie directed by Billy Wilder with a plot by Gene Weingarten featuring a young Al Pacino in the lead role and set in Brooklyn (okay sorry, the Bronx). You got it. How about sequels to "Wuthering Heights" or "War and Peace" by Brontë and Tolstoy themselves? Okay, forget "War and Peace." As the Wise Man of the Bronx points out we're talking about "ideas" --- whether expressed as characters, plots, images, brush strokes or a specific arrangement of notes, performed in a certain way. What it will boil down to is whether AI can get those "ideas" straight enough, at least to the point where you are able to suspend disbelief sufficiently to believe you are reading an original or seeing a professional presentation. Or whether you care. It's not a matter of forensic analysis afterall, it's for enjoyment. Beethoven's "10th" is a case in point. It sounds like Beethoven, but to the trained ear of the aficionado or musician it is missing that "special sauce." Yet it is close enough to simply enjoy in the absence of any more original works to come for a lover of Beethoven.
There was an excellent segment on this past weekend's On The Media with a composer talking about AI in the music field: https://www.wnycstudios.org/podcasts/otm/articles/how-ai-and-algorithms-are-transforming-music
Looks like the WaPo brain drain continues with political correspondents Ashley Parker and Michael Scherer reportedly leaving to join The Atlantic. Josh Dawsey is also said to be on his way out or at least seriously considering an exit.
I have seen a number of works done in the style of another artist. (Jack Davis of Mad Magazine fame seems particularly popular right now.) Is this any more theft than using AI?
The “borrower” in these cases like Mad usually sign their names. Ai seems not to ever do so. (I’m now using the small “i” so I stop reading that someone named Albert is ruining our lives.)
The genie is out of the bottle. We can wring our hands but AI is here to stay. Legislative guardrails will not have much effect.
As we’ve already seen in political campaigns, the use of AI, especially photos, will result in ever more hard-to-disprove disinformation/lies. How will it affect the admissibility of photos as evidence in trials?
What would an AI stairway drawing in the style of MC Escher look like?
AI used to mean Artificial Insemination, like with dogs and bulls.
That seems rather pointed--unless Gene has spotted his error and corrected it.