Generative and a “re-generative” world

Once upon a time I was paid to write software.

It was momentous. I was being paid real money far in excess of what I made on my paper route but more importantly, it had been love at first sight.

I fell into it by accident. As a kid that witnessed Gemini, Mercury and Apollo in the moment, I not only unequivocally knew Grissom, Glen and Armstrong to be the stuff of heroes but also as neither an American and a kid with glasses, the only real option to participate was to become an astrophysicist.

The plan was looking good, I had a place to read Physics at a British red brick but it become unstuck having done what many have done and taken a gap year. Away from the bosom of home and fending for myself in Paris, I quickly learnt about life in crappy digs and eating stale baguettes dipped in black coffee. Money can’t buy happiness but it can make life more comfortable and with that the thoughts of astrophysics were discussed with my tutor and we decided that Computer Science as an elective might assail this young ambiate.

Of Automata and Simula 67

It was a thorough grounding. We seriously wandered around block structured languages, even delving into the practicalities of Simula 67, the progenitor of object oriented C++. We wandered through the arcane world of machine assembler, the closest you can get to bare metal. True to the British educational ethos, we’re thoroughly grounded in theory to underpin the practical and I become imbued in automata theory, predicate calculus, formal grammar and more theorems than I ever imagined existed. Turing, Chebyshev, Chomsky, Green and Shannon ring loud decades later.

I was home, I had arrived and never, ever looked back.

Someone decided that the tech world would be better off if I was a manager, then a director and often a VP. Punishing, exciting, rewarding and maddening, the decades as a paid participant have given way to a gentler existence as an active bystander.

Don’t bother me, I have the COVID blues

Wallowing in the sorrows of Covid, I marveled but ignored ChatGPT’s arrival. Stunningly heady adoption rates but hey, I’m on a multi year radio bender and surely this is just another hype moment that tech does so well.

Months later, I read with disbelief and horror that somehow large language models had been coaxed into creating usable software. How this uniquely human and throughly cerebral process of writing software could be done by a computer system both disturbed me and simultaneously brought home the true promise of a “better world” powered by artificial intelligence.

Too loud to ignore

I’ve spent the summer seeing how my old skill is now redundant. I’ve had many a journey with Google’s Colab, Gemini CLI.

ChatGPT is keener, happy to go the extra mile and seems to have learnt about me and my interests. Prolog, a post Red Brick event, has always tickled my fancy and ChatGPT spews out supposedly usable Prolog to construct a simulation of a hex board game that lionizes past battles. Sometimes, I wondered if my exuberant summer use of Google’s Colab had placed this random Californian at the top of a list of SOTA and POTA API abusers.

Node Red is dead?

This all came on the back of an immersive journey to learn Node Red, a graphical and effect language to construct applications either quickly for the initiated or simply be a lower barrier for the uninitiated. The end goal was a Node Red program to assist me in finding ham radio contacts within the niche communities of SOTA and POTA.

Generative AI software creation, as real as it gets

Feeding my Node Red program into Gemini and ChatGPT, I received a series of attaboys. Flattered and somewhat overexubernt, I let Gemini have at it to create a REACT/Python equivalent. It all looked promising but the wheals came off hours later as we found ourselves up a gum tree and me lacking the energy to see how to save Gemini from a far too expansive brief.

Double checking, Gemini accepted that all its work was being consigned to the bit bucket. Frustrated, hundreds of thousands of tokens consumed and real dollars being taken from me, I did what I should have done and had learnt over the summer.

Steps by step, I described in plan English what I wanted. Almost like a rose bud opening into a fragrant gem, my idea took shape. It’s simply another variant of my Node Red program, just simpler and no logging. Don’t like the text in a button, just say that in English and Gemini jockeys the REACT code to be what I want. Merge SPOTs that suggest the presence of a ham on a frequency, no problem, ask in English and magic happens and Python appears. It works. Step by step I add functionality. Filter the merged list, just Morse Code please, done! Filter the list, just 40m please, done, and so this goes, crafting something that matches my idea of how I will sit on a peak and turn awareness of fellow hams operating into what? Completing an award such as ARRl WAS, yes. Cranking up my Summit to Summit tally, yes.

It’s actually quite exhilarating to have created something in a few hours. Node Red took weeks. Maybe my software engineering skills aren’t entirely redundant. Clearly, I didn’t code the final product but I imagined something useful, I decomposed it’s creation through a series of English instructions that always imagined how I would write and debug the code, small, manageable bit sized functionality. Test it, refine, move on.

Satisfied, I take Gemini’s output for a spin on my favorite local “radio” peak. I love it, no surprise it works the way I like to hunt. That is a reward in of itself but as a bonus I nab a UT CW contact (Forrest, KI7QCF), one more of the remaining nine states to get ARRL WAS Triple play. Well done me, well done Gemini.

Wherever you may be on the AI spectrum from luddite to deranged cheerleader, something quite profound has happened in my much loved world of software engineering. It’s not an evolutionary change but rather a tectonic moment of destruction. It’s ironic that the software engineering world that is creating AI systems is then using them to creatively destroy a big portion of their own world. Kinda weird actually and maybe it’s capitalism at its most efficient.

I’m glad to be retired and an informed voyeaur. The software engineering world was always stressful but I suspect the stress level is in the red region now.

Long live change!