Move 37: the path to irrelevance

I hadn’t planned to end up in a shopping centre that Saturday.

I was in Pescara, where I’d arranged to meet someone to buy a skateboard, but just before leaving I noticed a small event going on inside: the Fun Fest.

I pay, say goodbye and, curious, decide to leave my new board in the car.

My partner and I head into the festival: it was packed with stands dedicated to Pokémon, Magic, World of Warcraft and the world of comics.

Walking between the stalls, my attention was caught by an artist who was drawing someone as a superhero. Watching him, I thought about how much time a job like that takes, and I wondered: I bet that with a well-written prompt ChatGPT could produce a similar drawing with an acceptable result, if not a better one, compared to what the artist was creating with so much skill and patience.

I felt almost dirty thinking that, like a cheat. I looked at that talent, that passion and that patience as something replaceable, no longer essential.

But there it is.

It’s not the first time I’ve had thoughts like these, because I usually have them about myself. I picture the work I do being slowly drained of the creative part I used to love so much.

For a while now, before doing anything potentially creative in my work, I turn to ChatGPT first. Sure: I rework the output, I adjust course by giving it more precise prompts, but it did the bulk of the work.

That detail which once would have come from my experience alone (and all of StackOverflow) and my own ingenuity is now done by ChatGPT in my place.

It robbed me of that hit of dopamine that would have made me proud, something I would have shown off about to my dev friends in the group chat.

And if the tone already sounds pessimistic, the next lines might be even worse.

Not long ago people said that AI would never be able to match human creativity and that it would probably just copy rather than create, because — so the argument went — it lacks the human components that set us apart: emotions and subjective experience.

It’s worth remembering, though, that AI, while unable to feel emotions, is fed by an ocean of information: that’s why it can spot patterns we couldn’t even see and, in the end, devise new strategies and create original combinations. In other words, AI has proven to be creative, but in a different way from us: it has an alien creativity.

I recently read a story in Nexus by Yuval Noah Harari that struck me: it involves Go (a hugely popular game of Chinese origin, far more complex than chess), a world champion of the game and an AI.

In short: there was a Go match between the reigning champion and the AI (called AlphaGo), and the AI started making a series of moves that the commentators (professional players themselves) didn’t hold back from sarcastically calling “anomalous”, moves so senseless in their eyes that they joked about them. The smiles faded, though, when the champion started to get into trouble. Several times. Eventually losing the match 4 to 1.

AlphaGo had come up with a whole new strategy, what is now known as move 37.

Isn’t that creativity? Or is it just mathematical patterns? And in the end, what is creativity anyway?

What will become of us, now that we’re no longer beings of exceptional creativity?

In an earlier period, I tried to convince myself that AI would be the tool that improves our working lives by expanding our abilities and boosting our productivity. But what will become of us when our abilities are no longer needed, because a barely detailed prompt will be enough to write what today takes us months, years? Sticking to my own field, AI tools like ChatGPT devour an incalculable amount of data, and when I think of GitHub, whose Copilot scrapes who knows how many code repositories (they say only public ones), I wonder: how long before AI learns to write its own code based solely on the requirements given by whoever it’s talking to? What will become of all those “good code” practices when they turn into a completely irrelevant implementation detail?

Maybe in the near future we won’t see mass layoffs replacing every developer with an AI agent. But I’d like to put forward a line of thought.

I spent five years in consulting, on brutal projects and in teams that were sometimes very large. Some were marked by deep inefficiency, others by a massive dose of grunt work.

The team makeup was usually the same: a few seniors, some mid-levels and lots of juniors.

And it was the juniors who were often the problem: they weren’t very autonomous, mentoring them was hard, and reviewing their code even harder.

Now imagine if the cheap grunt work done by juniors were handled by a single AI agent. Sure, it might need some course correcting, and reviews will always be needed, but you won’t have to explain that console.log('foo') shouldn’t be acceptable in production, or that swearing in code comments is best avoided.

And the bar will keep rising; it won’t stop. Today it’s the juniors, tomorrow the mid-levels, maybe one senior will be enough.

If in the code editor things seem to be turning in AI’s favour, on social media the technical discussions among people in the industry have got worse. More and more people are giving up on working through ideas or writing in their own words, settling for a prompt or a handful of notes to produce entirely AI-generated content. The result? Total flattening, where every piece of content seems written by the same person. Where’s the human element?

In cases like this, I don’t see any avant-garde, just a sad surrender.

A passage I read in Purple Cow came back to me while I was thinking about this new reality of AI-generated content.

The author describes how, on a motorway trip, he was enchanted by the sight of cows grazing in the countryside, but after a while the wonder turned into boredom.

[…] I remember that, driving along the motorway, we were enchanted by the sight of […] hundreds of cows grazing in the gently rolling countryside, in a setting that looked like something out of a storybook. For dozens of kilometres we kept gazing out of the windows in wonder at the beauty of the landscape around us. After twenty minutes or so, though, we began to ignore the cows. The ones in front of us were just like the ones we had just seen, and what had amazed us at first had become completely normal — worse still, boring. […] They may be perfect, well-shaped […], but they are always boring.

Just like those cows, AI-generated content, however perfect and flawless, ends up lacking that spark of originality that makes a human thought unique and imperfect.

I don’t want to be a hypocrite about this: the texts I write — like this one — I run through ChatGPT, but with a strict rule I set for myself: “only fix grammar mistakes, typos and punctuation, and tell me if anything doesn’t flow”.

But in the end, even this is a small defeat.

And, deep down, I fear it won’t be the last.

AI will seep into our most intimate spheres, influence our choices ever more strongly and, all things considered, we’ll welcome it, because it will always be the right thing to do.

Our creativity, our thoughts, our abilities will soon all be delegated to an AI.

Move 37 isn’t just a triumph for AI: it’s humanity’s relentless march towards irrelevance.

Sources and further reading

• AlphaGo and “Move 37”

  • Documentary: AlphaGo (2017)

  • Wikipedia: AlphaGo versus Lee Sedol

• The hypothesis of human creative superiority

  • Yuval Noah Harari, Homo Deus (2015)

• GitHub Copilot and code automation

  • GitHub Copilot: Official documentation

• Training on internet-scale data

  • OpenAI, GPT-4 Technical Report (2023)