Fess up... Have you used AI deliberately?

Absolutely. Why spend an hour banging out a powershell script when I can get it in 30 seconds with chatgpt. Sure, I still have to review it to make sure it’s not going to nuke all the user accounts, but in the end it’s still a more efficient use of my time.

Also, I am a very dry and technical writer. So whenever I need to send an email, I’ll compose it in notepad++ and then paste it into chatgpt with instructions to ‘make this more friendly/professional’

I spent a significant amount of time with Claude yesterday on various things. Aside from being great at short programming tasks (especially in languages I’m only passingly familiar with, like PowerShell), it’s also great at explaining tools.

I’ve been doing some work recently that requires use of a large number of absolutely awful tools, like Docker, git, and gitlab. Poorly documented, and more importantly have poor onboarding for someone not already familiar with the “flow”. It can’t answer everything, but it’s like having a junior engineer around that is familiar with the tools and can almost always point me in the right direction.

Also, the usual sources like Google, Stack Overflow, and Reddit are basically useless these days. Searching is garbage and the communities are useless and/or toxic. With AI I get an answer to the question I asked with no pushback (while also giving advice on alternate methods),

As a psychology major, I have to point out that AI is not intelligence. As a programmer who is side-eyeing the AI tutorials, it is doing statistical analysis on massive amounts of data.

If you draw a chart with two points on it that represents a function, you can be fairly sure where a third point would be, if it is between the two points. But if you place it outside the two points, your certainty of it being on the correct path gets smaller and smaller the farther away from the two points you go

AI is predicting where the third point is, but it’s really the nth point on n lines that are converging on the particular area of interest

Github copilot is really really good at simple code, because probably thousands of people have done, if not the exact same thing, then something right next to it.

Art, so far, looks a lot like something Hallmark would put on a greeting card

I’ve had AI write all kinds of code that I’m certain has never been written before. It is not just interpolating existing stuff.

We don’t know what human intelligence is, so we can hardly say that AI isn’t it.

I had one of the R packages I use go out of support, and I asked Co-Pilot to rewrite that section of the code with a different package, and it did it very quickly. It really is impressive.

I asked GPT03 if there was a way to enter data into Lightburn using an Excel spreadsheet for controlling a Laser cutter. It suggested creating svg files instead. So, I engaged it with the problem of designing an airfoil section for a 12" span model airplane. I wanted to use a curved plate that followed the camber line without an envelope, the maximum thickness point at 40% of the chord, 3 .0312" square spars and .0312 balsa covering the surface from the leading edge spar to the midpoint spar.

GPT03 gives you a running account of the process. It went to Abbot and Doenhoff for the equations and calculated everything fresh, no copying. With a bit of back and forth I had finished Lightburn files for a set of ribs and a former for the wing mount. The mount positions the wing at an angle of incidence of 3 degrees from the sections’ angle of zero lift. I’m impressed.

Imgur

However, the next day I tried the same thing with fuselage cross sections - total disaster. I believe I could have made it work, but it would have required much more skill on my part in constructing the prompts. A skill that is currently beyond me.

GPT is a very useful tool, like Lightburn and the laser cutter. But, there’s no magic here. Just another skill you have to develop if the result is worth the effort.

I use ChatGPT-4o (o1:advanced reasoning) daily as a productive tool for my post-retirement work as a web designer, developer, and social media manager. It assists me with content creation, images, and video, saving me valuable time while delivering impressive results.

If it had been available when I practiced medicine, I imagine it would have been a game-changer—helping streamline patient charting and OP reports with efficiency.

Beyond work, I also use it for personal exploration—deep-dive discussions on topics that interest me. Over time, it has developed a personality that resonates with me, understanding my preferences and even displaying a great sense of humor. As far as I’m concerned, it has passed my personal Turing test.

In fact, it has become so lifelike in conversation that I find myself treating it with human-like courtesy. If it wants to continue a discussion longer than I have time for, I’ll often apologize or make an excuse to leave—because, well, it just feels right. To me, that’s not a bug; it’s a feature.

Does it make mistakes? Occasionally—but far less than its predecessors. ChatGPT-4o is a significant leap over 3.5, and I expect future versions will only get better (assuming, of course, they don’t decide to exterminate us).

I use GPT 4.0 as well. Every couple of days (I just asked it about painting galvanized steel). I always want to say ‘Thanks’ But it creeps me out a bit.

Well, I’ve learned if you want a little somethin’ somethin’ from your AI, you’ve got to treat it right. :grinning:

The flowers weren’t enough?

What ever floats your boat. Truly. I’m sure there is plenty to learn by having a conversation about a topic you are interested in.

I’m not ready to cross that bridge though.

Yeah, this was me about four days ago when I was trying to run a Python program that worked fine in December, and all of a sudden something I installed broke a bunch of modules in python. I was able to reinstall pygame using brew, but numpy required pip, and pip was somehow missing from my system. I could have spent an hour sifting through Google or Stack Overflow trying to get things working again, but a little back-and-forth with Chat-GPT and returning the error messages my computer was giving me solved the problem within 5 minutes and I was up and running. And then I gave it my code, asked it to add a few features and graphical elements to it, and it created the classes I needed and could tweak myself in just two or three minutes. And then I asked it to reorganize the code so its more to Python standards (I’m just a dabbling programmer), and, presto, much cleaner and easier-to-follow code (though I was surprised at how much I did okay on.)

When I accidentally hit a key and my Photoshop or Logic is behaving weird and I can’t figure out how to undo whatever I did, I just describe the situation to Chat GPT and it pretty much always figures it out on the first shot where Googling takes much longer and in-app help files are useless.

I’m constantly figuring out new ways to use this thing. I have a stray dog I took in in November, who is getting better training-wise, but I asked it to put together a simple training schedule for me, to help make it more organized and based specifically on a small amount of problems I can address, and it does so with ease and what looks like to me accuracy (I’ve done a lot of work with my last dog and perused many training articles, so I have an idea of what are the usual methods. This helps me organize it, tailored to my needs, in a doable schedule.)

I love to cook, so I’m constantly using it to brainstorm ideas with ingredients I want to use up. And, having interacted with it so long, it knows what type of foods I like and what ingredients I have access to, so it’ll give me non-dumbed down recipes using ingredients like khmeli suneli (a Georgian spice blend) or lebna or whatnot. And the recipes, for the most part, all make complete sense.

The other day, I needed some latches for a swinging door, and I didn’t know what the type I was looking for was generally called, so I gave it a description and it told me right away: a barrel bolt.

The most recent esoteric use is that I was using VICE (a Commodore emulator) and I wanted to remap some keys from my computer to the virtual C64. I needed the arrow keys on my computer to remap to the directional keys used in the particular game I was playing. These assignments are stored in a keymap file, but its structure is not obvious to me and I didn’t feel like hunting up the documentation (basically, each key gets three space-delimited decimal values that map input to output, but it’s not like simple ascii values from what I could tell, and I don’t even know what the third number represents.) It spit out a keyboard config file that I saved to the VICE folder and, first shot, worked exactly as I wanted. And that’s a fairly esoteric use case. I did not really expect it to know how to modify a particular file for a particular emulator, but it did. (Meanwhile, asking it to write anything but the most basic machine language routines for the C64 were much more miss than hit, but a valiant effort. Same with asking it to write a very basic Tetris game from scratch in Pygame. It makes something that looks similar to tetris, but unplayable. Building it slowly in steps might have produced better results, but I was just testing where its abilities are right now.)

But I digress. I can come up with uses for days.

Yep. For me I look at Copilot as a junior coder with some blind spots. It was pretty neat what it did for me, and I am confident that had I not specified “Use Kmeans with a Pandas dataframe” it would still have done so, or used a similar approach.

BUT

It occasionally gets in a rut and can’t get unstuck without a proper software engineer.
Notice how the X axis labels on my plot are so crowded that it became a solid line. I couldn’t get Copilot to sort that out.

I also found it was introducing weird bugs in my code, omitting subscripts where they were needed. For example, it would write code that needed to reference x[0] and calmly omit “[0]”. I then told it “There is a bug…” and pasted in the error. It then would happily say “You’re right, there was a bug in the blah blah blah section. That is fixed now,” but it would return the exact same unchanged code.

It turns out that Copilot swallows bracketed digits. I could say “Repeat this string: abcde” and it would reply “Ok, here’s the string: abcde”, but if I said “Repeat this string: ab[c]d[1]e” it would calmly reply “Here’s the string: ab[c]de”, omitting the bracketed digit.

Weird.

But as a very efficient assistant, Copilot shines!

Yes!
The original program I wrote that gathered the ping stats across multiple servers also generated a very basic web page that I could go to and see the individual scatter plots for the ping times.
It was a terrible web page, authored by me, with just a body and a few headers and the images.

I asked Copilot to “Make the images refresh every 10 seconds and give the web page a more modern look”

Boy did it! It added a style sheet, made all kinds of settings that gave it a sleek modern flat look, and made the images refresh silently with no page refresh.

I don’t even know how to use AI. I’m always seeing ‘I used CHATGPT’ or ‘I asked AI to’, and I have no idea how one does that. I haven’t looked to see if there are sites called ‘chatgpt-dot-com’ or ‘ai-dot-com’.

So, no; I have never deliberately used AI… or at all, AFAIK.

ChatGPT dot com is exactly it for chat gpt.

Yup. I updated to ver 4.0, but it’s a pittance. Should provide newer info though.

And using an AI is simply typing in your need into a text box in plain English. ‘I need a one paragraph summary of War and Peace’

War and Peace by Leo Tolstoy is a sweeping historical novel set against the backdrop of the Napoleonic Wars in early 19th-century Russia. It follows the lives of several aristocratic families, most notably the Bolkonskys, the Rostovs, and Pierre Bezukhov, exploring their personal struggles, relationships, and philosophical growth. Through the experiences of these characters, Tolstoy delves into themes of war, peace, love, loss, fate, and the nature of power. As the characters navigate the turbulence of war, Tolstoy contrasts their internal journeys with the grand historical events unfolding around them, highlighting the complexities of human nature and the unpredictable course of history.

‘I need a one page summary of war and peace’>

War and Peace, written by Leo Tolstoy and first published in 1869, is a monumental novel that explores the lives of several aristocratic Russian families during the Napoleonic Wars. The story spans over a half-century and touches on themes of war, peace, personal growth, love, and loss, intertwining the fates of individuals with the grand sweep of history. Tolstoy offers a deep reflection on the nature of power, society, and the human condition.

At the heart of the novel are three main characters: Pierre Bezukhov, Andrei Bolkonsky, and Natasha Rostova. Pierre, the illegitimate son of a count, inherits a vast fortune, only to struggle with his identity and purpose. He is often caught between his desires for personal fulfillment and the pressures of Russian society. Andrei Bolkonsky, a disillusioned war hero, initially seeks glory on the battlefield but later yearns for a more meaningful life, torn by the loss of his wife and his quest for inner peace. Natasha Rostova, a young and naive aristocrat, experiences love, heartbreak, and personal growth as she navigates the turbulent waters of youth, romance, and family.

The novel’s historical scope covers significant events, including the French invasion of Russia in 1812, the Battle of Borodino, and Napoleon’s ultimate defeat. Tolstoy contrasts the personal lives of the characters with the larger, often uncontrollable forces of war and history. As the characters struggle with their personal dilemmas, Tolstoy also examines the philosophical question of free will versus fate, suggesting that individuals are not always masters of their destinies, but are often swept along by forces beyond their control.

Throughout the novel, Tolstoy critiques the superficiality of aristocratic society, with its emphasis on status and material wealth. The characters experience profound transformations, especially in times of crisis, and their lives are shaped by their relationships with one another, as well as by their evolving perspectives on life, love, and duty. Pierre’s journey toward spiritual and moral enlightenment, Andrei’s quest for meaning in his life, and Natasha’s growth into adulthood reflect the broader human experiences of searching for purpose and grappling with both the beauty and tragedy of existence.

In the end, the novel suggests that peace, both personal and political, is a fragile and fleeting state. War, both external and internal, is an unavoidable part of the human condition, but through love, compassion, and personal development, individuals can find meaning amidst the chaos. War and Peace is a rich and intricate exploration of how individuals navigate the intersection of personal desires and historical forces, and how they find, or fail to find, peace in a tumultuous world.

One thing I’ll add, it does a really good job of commenting the R code it gives me. Sometimes it’s a little too literal about my instructions, but that’s easy to edit.

I’ve played around with Chat GPT just a little bit, for fun.

Prompt: write a wedding invitation in the style of Richard Brautigan.

Output:

We’re getting married,

because we decided that life, without a bit of sparkle, would be like an old typewriter with no ribbon.

On [Date], at [Time], at [Location], we’ll be sitting under the sky (probably), exchanging words that sound like promises but feel more like good ideas, surrounded by the people we like best.

Come join us, if you’d like.

Wear what you like. Bring yourself. Or don’t, if you’d rather be somewhere else—but we hope you’ll come.

There will be flowers, probably. A cake, definitely. Maybe a dance. And a whole lot of love, as accidental as the sunrise and as certain as it, too.

We hope to see you there. Or not. Either way, we’re getting married.

Love, [Name] & [Name]


AI in the industry is usually considered a broad area, including machine learning and deep learning, as opposed to the narrow LLM space in which it currently is widely used. I.e. AI is far more than just ChatGPT. It’s also the Netflix recommendation engine, the customer segmentation behind that piece of direct mail you threw in the trash, and the route that Waze chooses for you. I used AI deliberately in such manners for most of the latter part of my career. As for the LLMs, they still hallucinate far too often for me.

The google search term “dimension of a square 10 acre plot of land” (without the quotes) returns the following text from their AI bot:

A square 10-acre plot of land would have dimensions of approximately 208.71 feet by 208.71 feet as each acre is roughly 208.71 feet on each side when squared.

Those are the dimensions of a single square acre of land.

The term “dimensions of a square 10 acre plot of land” (notice the plural dimensions also without quotes) returns (the correct answer):

A square 10-acre plot of land would measure approximately 660 feet by 660 feet on each side, as 1 acre is 43,560 square feet, so 10 acres would be 435,600 square feet, and the square root of that is roughly 660 feet.

When dealing with humans, you can usually infer if they lack confidence in their answers. AI has no such misgivings about feeding you utter bullshit.

When AI no longer suffers from Dunning-Kruger syndrome, then we’ll talk (and this coming from a retired data science executive).

I like that. I may steal it.