Bell-Ringer Readings # 1
Selections from March-May 2025
From March
Item 1: He titles his post, "Why Can’t AI Generate A Glass Full of Wine," which struck me as seeking a solution to a non-existent problem (as opposed to a solution seeking a problem). The problematic AI-generated image showed a proper glass for red wine filled to the appropriate level, ie, where the glass is at its widest for maximum aeration surface area for aroma release—approximately half full. Untold hours were spent trying to force the generative engine to create an image of a glass of wine full almost to the brim.
Perhaps the actual problem was that, based on its training, the AI “knew” the level was correct, but the author didn’t. I don’t recall if he ever considered replacing ‘wine’ in the prompt with ‘water’ to see if the AI would then oblige. Diagnosis: (surprisingly) Good AI; Bad “prompt engineer.”
From April
Item 2: I found myself wondering if language is a technology. I wondered if others had the same thought. The answer was “yes” to both, particularly about written language, because it requires other, “unnatural” technologies. Some argued that speech was not because it was “natural” and the result of evolutionary processes. There were arguments found supporting each view. Argue the merits of each side among yourselves. I consider language a technology for reasons that will come up below in a different Item.
Language as a technology is not to be confused with Language Technology, which, according to the Wikipedia entry, “studies methods of how computer programs or electronic devices can analyze, produce, modify or respond to human texts and speech.” This is computational linguistics.1
Item 3: Samsung allowed employees to use Chat GPT, until three weeks later, when proprietary company information was found to have been uploaded to the bot, which can be reviewed by AI trainers at any time “to improve their system.” Which is worse: Yet another damning concern about AI, or increasingly evident modern employee/management ignorance and stupidity?
Item 4: From the ‘Ironic Rules for Thee but Not for Me’ department:
Quoted from the Terms of Use:
META AI released its new Segment Anything 1 Billion (SA-1B) masking dataset:
Prohibited Conduct
You may not access or use, or attempt to access or use, the Website (including the Segment Anything Materials) to take any action that could harm us, other Website users or any other third party, interfere with the operation of the Website, or use the Website in a manner that violates any laws. For example, and without limitation, you may not: …
…engage in unauthorized spidering, "scraping," or harvesting of content or information, or use any other unauthorized automated means to compile information from the Website; …
We may investigate and work with law enforcement authorities to prosecute users who violate applicable laws or regulations in connection with their use of the Website (including the Segment Anything Materials)
From May
Item 5: She asks, “Does adapting to change mean embracing change?” and, “How can we adapt to change when all our perceptions of it cause us to break out in hives?” The post was about the effects computer-based technologies have on institutions, society, the planet, and, especially, personal interaction. Although the post was not specifically about AI, I pondered why she also asked, “What is the point of change if it plunges us into the abyss?”
The post includes a quote from another source that uses another phrase that I see everywhere, that hits me like fingernails on a chalkboard: “the democratization of {insert subject here};” in this instance, ‘artificial intelligence.’ The insidious notion of ‘democratize’ deserves a vilifying post on its own, but for the purpose here, I will point out the difference between its intended wishful-idiocy and reality. What is happening couldn’t be less ‘democratic.’ What is happening is techno-oligarchic manipulation. Generative tools were given away for free for a trial period to assimilate into, and acculturate the masses with the technology via its public, whimsical products. They became so ubiquitous, embedded, and addictive that their removal would be social and political suicide, even if it were possible.
This was a market, legislative, and mass cultural endorphine carpet-bombing for the intellectually and creatively impaired, not anything to do with “democracy” —a politicized word now so deliberately broadened and bastardized it is just semantic slush. Who voted for this? If you deliberately choose to use these problematic, discretionary toys, then you did, by proxy, via your addiction to producing crap, and validated with your subscription dollars.
She (we) may not wish to jump on the AI bandwagon, but she/we never had a choice to opt in, and we do not have a choice to opt out. Never did; never will.
Item 6: Another post presents with the expansive title, Sure, Generative AI Will Never Replace Human Creativity. But It Might Just Kill It.
There is the obligatory discussion about job displacement and elimination, the testing of AI generative tools, but also a good paragraph discussing machine learning and alleged “creative capacity,” by quoting Hayao Miyazaki’s comment that AI-generated animation was “an insult to life itself,” after finding his style had been AI-pirated and parroted.
The author restates an observation that much of what is considered creative, be it human or machine, can be reduced to math and later computationally reconstituted. This was meant to stimulate a discussion on defining what constitutes creativity, and rightly so.
I praise the author for the following:
A friend of mine, an English major, once said that no two words have exactly the same meaning. They can be incredibly close, but are never exactly the same. That specificity and nuance is important at this juncture, because while most of us use these words interchangeably, their etymology, their source, their history has an impact on their effect.
For those of you who pay attention to how much time I am spending in the AIgitated project on lexical issues and the problems they cause in historical discussions, you know this is a big deal with me. I put this problem on par with any historical technical issues. Language is a technology because it is a mode of transmission for intellectual blueprints—ideas. Starting with post 2.4 in AIgitated, I will be providing examples to drive this point home.
Item 7: The title asks: How Much Retouching Is OK? This is a common issue across many photography forums and probably the most uncomfortable and contentious area of debate. It is a troubling gray area that, contrary to much opinion, is not strictly a digital issue. But this post took a different approach.
He discussed a theoretical lecture he attended years ago that attempts to deal with this problem as an ethics issue by defining “spaces,” ie, descriptors of the type of photographic purpose (generalizing as documentary vs artistic) and what can or cannot be ethically altered within each defined “space.” These points could be endlessly argued back and forth. The attempt to define a practical boundary for ethical editing in this way is elusive and, I think, impossible, because it does not fully frame the issue.
It misses the aesthetics, which is why the eye detects anomalies almost instantly as something being “off,” even if the proposed “spaces” rules were followed. It is subconsciously recognized and can be difficult to express in words. It’s not necessarily deceptive, but unnatural in a way that deliberately abstract works are not. It can be subtle, but it is obvious, akin to Justice Potter Stewart’s assessment that he can’t define obscenity but knows it when he sees it. It is the reason even slightly over-processed images are instantly unsettling. Our subconscious recognizes something outside of its innate perception of reality.
I have argued my perspective on photography sites for several years. My take was generally not well received because photographers would not face up to their addiction to the sliders and tools in their editing apps. They do something because the tool allows it to be done, rather than asking if it should be done.2
There is growing awareness of the gratuitous ploy of marketing everything possible as being allegedly “AI-powered.” AI has always been behind all the sliders and tools in digital editing apps since the beginning of the digital age. Not only is the marketing hype now on steroids, but so are the apps. I am slowly seeing a shift toward my views on these issues after holding up a mirror over the last two and a half years, as I attempt to persuade photographers to deal with this issue in the face of this new AI onslaught, now.3
This is not an issue just for the “photographer” class with their cameras, gear, and expensive editors, portfolio sites, or blogs. With cellphones, “everybody is a photographer,” and these concerns, and others, are things “everybody” needs to think about (Ah!…the ruckus I can cause over how your phone does what it does and what that should or should not be called).
I’ll end this installment here and pick up next time with items selected from my burgeoning database that were published in June. And things I am reading today have some beauts for a future post.
Anything referenced on Wikipedia is always a crap shoot concerning the validity of citation source and veracity. The all-too-often seen phrase, “according to Wikipedia,” is a statement that should be banned. Wikipedia is not the source, only an uncurated and unvetted, often badly quoted or paraphrased, sometimes correctly footnoted repository of other sources. Here, the sentence quoted was said to be sourced from the German Research Center for Artificial Intelligence. I did not find it on the institutional site, but on the document author’s site. What was quoted on Wikipedia, surprisingly, was correctly copied from the first line of the document.
Ex: Those obnoxious images of waterfalls that look something like frozen milk, not water in motion. no, No, NO…STOP IT! THIS is what water in motion looks like:
Deliberately coincidentally, all this coincides with the period during which I have not done any shooting because I spend all my available time working on these two projects. I can make time to do one or the other, not both (nor other activities I miss). “This” is more important.
** The spillway at Davidson’s Mill Pond, North Brunswick, NJ, on a foggy autumn morning in 2022. One of my last images before putting aside my cameras for these projects.