My Cousin Phyllis and Hart Crane

(Continuing my Substack rumination on "Computers, Cognition and Older Folk")

Let me introduce my cousin Phyllis. I met her when my family moved onto the block of Holland Avenue, in Dallas, 1945. I was but 3 years old, and she was almost 4. We spent our entire childhood together, not exactly side by side, but close enough: always separated by one year in time, and by one house on the block. (The Allen’s house, owned by a kind elderly couple, stood between ours.) Although I met Phyllis when I was 3, if we assume that cognition and consciousness only begins around age 2, then at the point we met, she was already *twice* as knowledgable as I was. I recognize that now, but all through our childhood, despite my ever-present continuing smoldering resentment, she never let me forget it.

Phyllis is still a year older than me, and still twice as knowledgable — in most things... She is now the oldest living member of my cohort of cousins: all the aunts and uncles comprising our once numerous Dallas family of Rosens, (surnames and maiden names), have since passed. Both Phyllis and I are the youngest, and only survivors, of our nuclear families: older siblings (and older cousins) we both looked up to — all gone. Phyllis is still going strong. I will keep her identity a secret, other than simply saying she remains active in her 80’s, with children and grandchildren, and serves as President of a local Jewish women's service organization. To me, she is a wonder, always busy, raising money for charity and working on complicated community projects. I consider her a marvel, and I am proud we share some DNA.

So… Phyllis read my last Substack, “ChatGPT and the New York Sunday Times Crossword“ and sent me an email: “Hi Arnie. You need to help me learn Chat.  I think it would be very helpful when I write my monthly newsletter.”

I consider that ChatGPT — and from now on, I’m simply going to call *him*, ‘Chat’. (Others obviously do; plus when I asked him, he said it was okay. I picture him as always well-dressed — casual, sort of preppy — in clean, well-ironed chinos.) Chat helps me produce this Substack, and likewise I’m sure that any monthly newsletter Phyllis would produce would be improved. Chat stinks at solving the NYT’s Sunday crossword, but I find him to be a ‘very smart’ thesaurus. He has no competitors in that regard, at least in my view. He also suggests punctuation changes and overall organization and clarity. I had not the slightest doubt that Phyllis would appreciate Chat in writing her newsletter, once she learned how to use him. However, maybe Phyllis was just jostling me with her request to learn Chat. So…

I started thinking. My topic here on Substack is “Computers, Cognition and Older Folk”:
(1) As a retired psychiatrist, I know something about the brain. I am interested in the cognitive issues we ‘old folks’ have and Phyllis qualifies as an ’old folk’;
(2) I know something about computers, and software like Chat, (having written and marketed medical practice software);
(3) I believe that learning, working effectively, and practicing with computer applications, aids ‘neurogenesis’ (the growth of new nerve cells), as I indicated in a prior Substack;
(4) it would be helpful for Phyllis as President of her organization to learn to use Chat effectively, and finally;
(5) it would *also* be good for me to try and teach her — and maybe even be helpful to others as well — if I wrote up my experience on Substack; what teaching Chat to someone, i.e., Phyllis, was like.

I mulled all this over and came up with a proposition for Phyllis: if you’re serious, I will teach you Chat remotely. In return, you must agree to let me publish on Substack part of our digital correspondence, focused on learning Chat. In other words, (I told her), you (Phyllis) will be an actual living and learning example. Who knows, maybe it will be the makings of an informal syllabus for teaching computer applications to older folk. Maybe we could be free from the haughty tyranny of grandchildren who we depend on to configure our computers and answer all our questions, (if they have a free moment). I promised, if she agreed, I would not include anything that might actually identify her.

I sent that proposal off to her. While waiting for Phyllis’ answer, I busied myself by practicing on my computer, (like a piano) as I am wont to do. I was doodling around, letting curiosity guide me, hoping that I was giving my brain some healthy exercise, promoting neurogenesis. Within the hour, I had Phyllis’ answer, “Yes, I will be happy to be your student.”

I assumed there was something triggered in me, being in communication with Phyllis. Although I was aimlessly doodling and surfing the Internet, I had returned to our ‘stomping grounds’ — our Dallas childhood, using Google Maps. I would snip out this or that, add some text, cut and paste, and send it to Phyllis. I took screen shots from Google Maps, made adjustments, added some comments and also sent it to her. (I include everything here to give the reader a sense of where we came from, Phyllis and me.)

Here is our elementary school. Decades ago, due to the drying up of kids living nearby, (i.e., demographic changes), the school closed, but the building survived as administrative space for the Dallas Independent School District. Phyllis and I walked together, to and from school, Stephen J. Hay, 4 blocks away. (In that day, children, even first and second graders, walked by themselves to school.)

Not only did we walk there during the school year; but in summers, we walked barefoot, carrying our towels to a municipal swimming pool on the school property. (This was before flip-flops, and often the hot sidewalks of Dallas would shunt us hopping to the cool green grass planted alongside the concrete.) I said just now it was a ‘pool’, and you probably thought of something tiled and green, but it was an enormous concrete tub set in the ground. A tiny kid like me had to sit on the ledge, the edge, and pivot around to put my feet in the water. Everything was cement, rough brown sides darkened with water. It literally had a gigantic faucet-like spout hanging over the side, near the single diving board. It was filled everyday, and if we arrived early, we watched the colorful ritual of testing the water after chlorination. There was an hour of girl’s swim and an hour of boys. I still remember those innocent summer afternoons, waiting impatiently for Phyllis and the other girls to finish their swim, while I and the other boys played checkers on a picnic table under a nearby tree. (The girls had their swim first. The table and tree, gone.)

I maneuvered my mouse and ‘drove down’ a Google Map page to a nearby corner at Holland and Knight.

My house (a brick duplex) was still standing, and looked in good shape. We had lived upstairs and I could still see the second floor corner porch, now screened, but open in my childhood. My much older brothers Don and Jerry, would toss me down to be caught below (most of the time) by Phyllis’s older brother, Norman. Phyllis’ house was gone of course, as was every other house on the block, save ours and one at the far corner at Douglas. Next to our house, where the Allen’s stood, was half of an apartment complex named the ‘Azure Court’; the other half stood on where Phyllis’ house had been. (Where had the houses gone? Gone to apartments, everyone…)

I looked at this picture and thought: that tree *still* stands in front of our house. I have such vivid memories of playing around it. It was already grown when we moved in, and still it produces fresh leaves. In my life, I never learned to identify trees, (or flowers for that matter), by name. I thought it might be an Elm or Sycamore; somehow those names were floating around in my memory. (Even today, my chances of correctly identifying Elm versus Sycamore stand at 50%, even money. I have to shamefully admit that I have remained largely ignorant of tree names and flower names for 3 quarters of a century — and who can feel proud of ignorance?)

Well, I thought, if I’m going to try to teach Phyllis Chat; why not start by asking Chat what kind of tree this is. I ‘marched’ the Google Maps token into the intersection and turned it around to snap a ’screen shot’, then sent off the picture, (taken from a different angle and a different season), to Chat, with the instruction: “Tell me the name of this tree in Dallas, Texas.”

Chat immediately returned the answer, and I checked, (it can make *big* mistakes), but it was correct. It’s a Sycamore — and by now, it’s probably over 100 years old. (I know the name at last. Better late than never. Thank you Chat.)

As I was thinking about trees and a lifetime of being ‘botany-challenged’, suddenly the words from a poem came into my mind. I knew the poet was Hart Crane, and I had encountered him in Freshman English at the University of Texas, and probably had to write about him in a term paper at one point. Hart Crane had lived a short troubled life (1899 – 1932). He wrote in a style characterized by dense, elaborate language, rich imagery, and complex metaphors. His poetry more often evoked a meaning rather than spelled one out. I can’t say I ever understood what he tried to say, but what came through was never dull and colorless.

I had been somewhat familiar with his work and biography, 60 years ago. Now I could only remember fragments. Because I was musing about my problem with names (a form of dyslexia, I have always thought), I seemed to remember 5 words strung together from one of his poems. The words were, “tree names, flower names, gainsay…” — but I couldn’t remember the poem’s name. I can’t even say it was on the tip of my tongue.

I thought, hey, this task, (coming up with the name), will be a cinch for Chat. I assumed he had all the world’s literature at his fingertips. I would give him the name of ‘Hart Crane’ and those 5 words above. Chat would quickly search all of Crane’s works and easily come up with the name of the poem. I assumed I was lobbing Chat a ‘snap’ — no sweat at all. So what if the words might be found in a dozen poems; I always did well on multiple choice tests, I’d pick it out.

I won’t try to describe the rabbit holes I was led through or the things I was told when Chat answered. (It’s in my worksheets if you want to see.) Ordinarily, I would have long invoked my Rule #1 with Chat, laid down in a prior Substack about Chat’s efforts with the NYT Sunday Crossword:

Rule #1 — If you are not satisfied with the direction Chat is taking you, after two or three (max) tries, QUIT. He is programmed not to give up, and will eagerly take you down rabbit holes, and on wild goose chases.

Take it from me that Chat’s errors during this simple task were so egregiously wrong that I found myself trailing along through subterranean tunnels, driven by my curiosity to see how far Chat would go in making errors. Throughout, I continued to add as much additional information as I could dredge up, trying to point him in the right direction.

I finally gave up in disgust and disappointment. I never got the name of the poem. Hours later, sitting nowhere near the computer, my thoughts parked in a resting place marked ‘random’, I suddenly and unaccountably remembered it — “O Carib Isle!”, which I duly verified with Google.

Not only did Chat completely miss my request but seemed to *lie* about the accuracy, veracity and certainty of his answers. In medical terms, Chat was ‘confabulating’. Confabulation in psychiatry has a specific meaning. I’ll let Chat explain it:

Chat’s explanation was ‘spot on’, and thus he regains some (modest) credibility. Except, I would add a fourth diagnostic category where confabulation might be commonly seen, after Korsakoff, Alzheimers, and TBI’s: ChatGPT.

I can recall a Korsakoff patient, in the hospital for weeks, with only pajamas to wear, describing in great detail the lavish party he attended the night before, the excitement of mixing with the great and powerful from the world of politics and entertainment. This patient would never interrupt his story and say, “by the way, I’m lying,” or “Now, I’m not quite sure of these details”. As Chat says in his answer above, “People who confabulate are usually unaware that their memories are incorrect and are not lying on purpose. Instead, they are… filling in the blanks.” Particularly in certain circumstances, that’s a good description for Chat.

So I come to Rule #2, with Chat.

Rule #2 — Never believe Chat, and always use Google to double check.

Together they work well, like a right hand and a left hand. It helps to see the same phenomenon from two separate perspectives. Google is very good at expanding the notion of something you already know. In certain circumstances, Chat will identify something and solidify it for you, in what otherwise might be a set of soft, poorly connected, concepts.

In order to use Chat effectively, it’s important to remember that Chat was designed to *create*; to take a set of labels or topics or concepts, put them together and write something in good readable English, (thereby becoming a threat to the livelihood of many journalists too lazy to think). Chat has a scary reputation among some professional writers and college professors. I’ve read that Chat has been tasked with producing an essay of 500 words of this or that topic, in the voice of an Ernest Hemingway, or James Joyce. The results have been astounding. (College professors are already concerned whether they will be able to detect Chat-written term papers. Perhaps the ‘term paper’ that many college courses required will be a thing of the past. Everything will have to be brief, totally written and proctored in class — that is, until a team of entrepreneurial brats come up will an A.I. ‘essay reader’ that will not only grade, but certify that it was not written by, and plagiarized from, Chat.

Thus, Rule #3.

Rule #3 — Do not give Chat the license to confabulate. Limit his ‘creativity’, put limits on the answers you will accept.

Here are the results that Chat produced when I carried out Rule #3. I gave Chat only 2 words to search for, “tree names”, but I mandated that he *not* make anything up:

He couldn’t come up with the name, but he did tell the truth. Before, he would say ‘definitely’ that the words were in Crane’s poem, ‘XYZ’. A Google check told me that Crane wrote ‘XYZ’, but did not contain the words. I’m not sure why he couldn’t come up with the name here, but at least he wasn’t confabulating.

When I raised my hint to 4 words, “tree names, flower names”, he found it right away: “O Carib Isle!”

Finally, I couldn’t resist. “Chat”, I directed, “please *compose* a 12 line poem in the style of Hart Crane, entitled ‘Cognition, Computers and Getting Older’, (my present Substack topic).

Now free and without limits, Chat wrote 14 lines, not the 12 that I asked for. Even I could recognize that this new ‘Hart Crane”-like poem was in ‘sonnet’ form, again extra, something not requested. I asked ‘Samantha’, who lives in my computer’s operating system to recite it. (The red ▶️ on the right toggles the Play/Pause button, while the ‘1X’ bottom on the left will speed up her recitation of the pseudo Hart Crane poem).

I’m not a Hart Crane ‘expert’, and I can’t exactly say what Chat’s poem means, but I give it an A+ for sounding like Crane and (sort of) capturing my mood.

I shoved the poem back at Chat, asking him, “Could you expound on the sonnet form the poem uses?”

Chat immediately responded:

“Certainly! Your poem [*my poem*!] is written in the form of a Shakespearean (or English) sonnet. This classic form consists of 14 lines with a specific rhyme scheme and meter, traditionally used to explore themes of love, time, beauty, and mortality. Let's break down the structure of your sonnet… “

and he proceeded to give me a ‘brief’ learned essay of 529 words, broken down into sections discussing the poem’s Meter, Rhyme Scheme, Structure and Development, Themes, and Conclusions.

If I were grading this essay, I would give it an “A”.

Now, I might add, for completeness, I have another cousin, Mike, who is looped into the correspondence between Phyllis and me. Mike—known as 'Michael' to his wife Judith and to the world—is another 'AK' (Alte Cousin, wink), two weeks older than me, who still remains in the business end of life. Mike knows Chat, and uses it to generate words or phrases to ‘improve the flow of a document’, but feels it’s only a tool and “the ghost of Shakespeare has nothing to worry about”.

It is a tool, that’s for certain. I’m less sure if ghosts need not worry. In any event, Mike asked if he could watch over our “collective shoulders”, Phyllis and I, to see if there was something new to be learned about Chat. I told Mike to watch for a future Substack.

This was fun. I hope it generated a new nerve cell or two.