(If you’re not yet a full subscriber to Herb On the Street and my Red Flag Alerts, and wonder why you should subscribe, I explain right here..)
The following is a fun freebie.
I’ve been playing with ChatGPT almost since the day it was rolled out...
I’ve found some things it’s good for, like summarizing a long story I’ve written, or converting it to bullet points.
Or helping me get over a writer’s block for a headline. (Truth be told, that’s where I got “Atomic Frenzy.”)
Or along with Google’s Gemini, speeding up my search process, with the full understanding that hallucinations, the term for false answers, are real and that you still have to double-check any and all answers.
I’ve seen white-labeled versions of gen-AI used in quite a few novel ways, such as Paragon Intel’s focus on ferreting out companies that are off-the-baseline evasive in answers on earnings calls.
On the Other Hand
There’s the downside, which I’ve seen first-hand…
Such as its use, in the name of efficiency, to replace the human touch on something like a press release.
Or its use, in the name of feigning intelligence, to answer a seemingly complex question.
Or its use to create images…
These are all easy to spot because they exude a mediocre and bland sameness that, at their core, lack the soul and vision that come with the human touch.
Guilty as Charged
Full disclosure, as you can see from the above image and a quick peek at my archives… I have been a big user of gen-AI for images that anchor my reports, essays and articles. I even pay for it, under the possibly false assumption that by doing so it works faster and the images are better.
Sadly, stock photos and graphic artists are simply too expensive for most independent newsletter writers. (If I’m wrong about that – please let me know.)
Free stock photos are often awful, and photos in the public domain – such as the one atop this piece – are limited.
That leaves AI-generated illustrations. Early on they were like a new toy and, at times, seemed to be a great hack to add an anchor photo.
‘That’s AI, Right?’
Then – like AI-generated writing – they started to look the same. “Oh,” my wife often says now, unimpressed, “that’s AI-generated, isn’t it?”
To be fair, it has created some I like, such as the image I currently use for my “Why Subscribe to Herb on the Street” pitch….
And as the illustration for “What to Expect from Red Flag Alerts”…
But there clearly are limits, especially when you have an idea in your head and try to tweak images that simply don’t cut it.
The Breaking Point
That’s when you realize: no matter what gen-AI supposedly will be able to do at some point in the future as it learns and re-learns, it simply can’t think.
I finally hit the breaking point a few weeks ago when I wrote an updated Red Flag Alert regarding Hims & Hers HIMS 0.00%↑ that was headlined, “The Weight Loss Illusion.”
For an anchor illustration, I fired up the ChatGPT image generator and did what I often do, starting with simple prompt. In this case it was…
An image that would go with a story that is headlined, "Weight-loss illusion."
This is what it spit out…
So I shot back…
No something more to the point - maybe at a fun house with mirrors.
This is what I received…
Worse, the illustration was accompanied by this message…
Here's the new image, set in a funhouse with distorted mirrors reflecting exaggerated body shapes to highlight the illusion of weight loss. The playful setting brings attention to the theme of distorted body perception in a more direct and whimsical way. Let me know if you'd like any further adjustments!
To which I responded…
No, it should show a heavy person who looks in the mirror and is skinny – but not overly heavy so as not to be offensive.
I went on to elaborate on my initial prompt…
I need an image that would go with a story on "the illusion of weight loss." Is there something that would show someone looking in a mirror that creates an illusion of being thinner than they really are?
And (drumroll!) I got this…
So I tried again…
How about someone at a fun house who is somewhat heavy but looks thin in the mirror.
This time I stressed that the person in the mirror should look thin. Pretty straight-forward, or so I thought. But – no. Instead it gave me this…
And to show just how poorly this was going, ChatGPT explained (my emphasis)…
Here's an image set in a funhouse, where a person who is slightly heavier is looking into a distorted mirror that makes them appear thinner, highlighting the illusion of weight loss. Let me know if you'd like any further changes!
It did just the opposite of what it said it did! Increasingly frustrated, I tried again, writing…
The image above shows the opposite.
To which ChatGPT responded…
It seems the image did not align with your request. I'll revise it so that the person in the funhouse appears heavier in reality but sees a much thinner reflection in the mirror, correcting the illusion. Let me adjust that for you!
Bingo! The heavy person should be looking into a mirror and see a thin image. Seems like it finally understood what I was going for, right? Wrong! Incredulously it gave me – this…
So, I tried one more time, patiently explaining…
No, the idea is that the person looking in the mirror is heavy and the mirror makes them look thin.
ChatGPT, this time very apologetic, said…
I see where you're going with this now! You want the mirror to create a dramatic slimming effect while the person remains heavier in reality, emphasizing that "illusion of weight loss." Let’s imagine the funhouse scene with a playful twist but keep the underlying theme serious—like those moments we all experience where perception doesn’t match reality. I’ll tweak the concept to really capture that, with the person looking heavier but seeing themselves surprisingly thin in the mirror.
Let me adjust the image once more and get it right!
Even after regurgitating my instructions, it gave me a choice of two images… both not even close.
And it must not have realized its mistake, because it went on to say…
Here's the new image, showing a heavier person standing in front of a funhouse mirror that makes them appear much thinner. This emphasizes the illusion of weight loss, with a clear contrast between reality and the reflection. I hope this aligns with what you had in mind!
This went on and on until finally I searched online and found a photo online from an old Glamour magazine and asked ChatGPT to replicate it…
Hey, how about something like this?
And it gave me this…
This continued several more times until I finally said…
No, never mind. You can’t figure this out.
To which it responded…
I apologize for not getting it right. I understand your frustration, and I appreciate your patience. If you ever decide to revisit this or need help with something else, feel free to reach out. Thank you for your time!
Spoken like a true chatbot.
Hal from “2001: A Space Odyssey” would be proud.
Left Brain vs Right
In the end, I realized the Glamour photo was so old I could probably use it, just citing the publication. That’s exactly what I did.
Maybe I just need to be schooled on how to put in the right prompts. Or maybe, just maybe, no matter what the hype is, when it comes to creativity, machines are like left brain people trying to operate in a right brain world. They simply don’t get it… and never will.
DISCLAIMER: This is solely my opinion based on my observations and interpretations of events, based on published facts and filings, and should not be construed as personal investment advice. (Because it isn’t!)
I can be reached at herb@herbgreenberg.com.
Funny, but the left brain/right brain idea has been debunked. Ask AI!
The clock in the early photo has 13 hours!