Lots of noise about the shortcomings of GPT and LLMs. But not from me.

Anyone who has spent time with our new friends has probably been, in equal parts, sceptical and excited by what’s happening — strewn with errors, false information, bias, canned, indistinct and lazy.

Like many programs and meetings, I’ve been in with human beings.

Nothing I’ve published (ever) has been written using Chat-GPT or similar AI. Not because it’s not helpful or I can’t get the result I need but because that’s not what’s valuable about it — to me. Naturally, we are all in very different contexts.

Partners In Thought

When explaining AI's current value, I think ‘thought partner’ — copilot to my riff. A well-connected creative ‘prompter’ able to get enough of a sense of where I’m aiming to suggest ideas. An obedient assistant to help me ‘engineer’ an outcome. A new kind of industrial magic. A tool for our age and one we can not dismiss. (I’m still deciding which)

And it’s teaching me to think.

It’s teaching me to create prompt strings to get what I want. Of course, that’s trial and error — but after a while, you find yourself thinking in the right terms —immediate feedback helps you modify/achieve your aim.

Everything Is Aiming

I read the other day that archery masters often teach that ‘everything is aiming.’ Where you place your feet, how you hold the bow, and how you breathe during the release of the arrow — all of which determine the result.

Awa Kenzo, the master archer, was so mindful of the process leading to an accurate shot that he could replicate the exact series of internal movements without seeing the external target. He could do it in the dark, blindfolded and right in the bullseye, with the second splitting the first.

Zanshin is the mind entirely focused on action — fixated on the task. Constantly aware of your body, mind, and surroundings without stressing yourself. It is an effortless vigilance. I love that.

Creativity Is A Process

You can engineer the result by refining your prompt (instruction) and ‘seeing’ the results — creative iteration.

It isn’t a simple act, and it takes time and iteration. It’s always about solving a problem. I’m OK with a blank sheet of paper, but that results in highly subjective effects. I’m the client, so I’m the only arbiter.

Problems come with a specification — the brief. If there isn’t one, then Job One is to create one.

A few things in life force you to think — alarming experiences, opportunities, people with challenging or opposed positions — a great story, perhaps. But very few cause you to think differently, especially in your own head.

AI As Engine And Engineer

The ‘engine’er’ as a term comes from the 14th Century. It meant someone who constructed military engines like catapults and other ‘siege engines.’

And before that, the word ‘engine’ comes from the Latin ‘ingenium’ (c. 1250), which means an ‘innate quality, especially mental power, hence a clever invention.’

Ignore/dismiss it, as many people will.

But as it evolves and reimagines how we think and work, it will find its rightful place in history — remembered rather like the wheel, the steam engine and the smartphone. If you can remember the dawn and enormous impact of the GUI? Well, a lot like that.

Visual Thinking

Regarding AI and beyond Chat-GPT, I am learning more about how to engineer a good result using text-to-visual AI rather than text-to-document.

I’ve done a lot with each big platform, some obscure ones and my clear preference is Midjourney. I’ve collected the baseline that works for me below.

It’s a rewarding experience having a vision/idea in my head and then prompting my ‘digital buddy’ to help me achieve it. And to repeat, I’m not doing this to replace my drawing by hand or create art in other forms but because it’s a fascinating and intellectual workflow.

I love the idea of being an engineer.

My father gave me that context in the ’60s.

I was a child in short trousers. But I could already admire his handwriting, ability to draw with both hands, imagination, and big-picture creativity. He sat at an enormous draughtsman’s desk with cantilevered lights and parallel motion drawing machines I dreamed of having one day. Not a computer screen in sight, but magic happened all the same.

He drew aircraft and rocket engines. And in those days, he did it all with a Rotring™ Pen, rulers, set squares, and french curves while smoking a pipe—an engineer.

Nowadays, we can speak to our computers, and bingo — nearly six decades later — out pops pretty spectacular art. Creativity that, while prompted by our brains and hands, we couldn’t make in 3 weeks, let alone 3 minutes.

My digital sidekick augments instead of replacing any part of my creative process.

Just Some Context

It’s still ‘early’ days even though, to most people, it happened in January. But it’s been 57 years.

ELIZA — the first of its kind chatbot and language model, was developed by Joseph Weizenbaum in 1966 at MIT.

In 1950, Alan Turing proposed the Turing Test. Experts in the field were animated. But truthfully, the only genuine interest in Artificial Intelligence (AI) was in science fiction.

Personal computers did not exist — the Kenbak-1 being the world’s first (c.John Blankenbaker) of Kenbak Corporation in 1970 — first sold in early 1971. Interactive computing was new, and natural language processing wasn’t well-known.

ELIZA must have seemed like magic to those in the know.

Prompt

Forcing yourself to ‘prompt’ for a result is different to many other experiences and, as such, has had a real impact on me. That brain-to-computer discipline is a fair bit dissimilar to any other thought process I can currently think of. Other than pure story-crafting and storytelling.

An assertive discipline — a skill needed for the near future.

Be as clear as you can about the result you want. Write it out and be as detailed as you can. Include many adjectives and be very deliberately instructive with the description.

Prompts — Here’s What I've Learned So Far

A possible structure for text to visual — Explain the style you are after — abstract, neue sachlichkeit, impressionism, baroque, photograph—lomo photography (there are thousands) — the camera angle, the specific kind of lighting you are after, the mood, the season and weather conditions, the type of sky and time of day, the particular objects you want in the scene, the degree of detail/resolution 4k, 8k — HD — and the emotion you wish to get across, and suggest an artist to evoke — Rene Magritte, Yayoi Kusama — Jean-Michel Basquiat for example, the texture, the aspect ration, the people, characters or species as required and the era or genre — cyberpunk, steampunk, meso and nanopunk, silkpunk and so on.

A possible structure for text to the article— First, sit back, relax, consider the high concept, and get clear on the topic. Then be clear on the focus — the topic, theme or big idea. Make that the main subject you want help writing. Explain the audience you are trying to speak to — suggest the tone and language. Explain what you are trying to achieve with the article — its purpose. For example — is it informative, newsworthy, educative, entertaining — and the tone — professional, light-hearted, witty, in the style of someone Ernest Hemingway for example — words like that. Explain how long you want the article to be—the keywords you wish to cover or for SEO you need. Add references, examples of proof or interest and perhaps quotations as sentences or paragraphs.

Thought Partner? Copilot? Creative Prompter?

Lots of noise about the shortcomings of GPT and LLMs. But not from me.