The first of a diptych that looks at my personal use of AI and LLMs.

At first, I did not have high expectations, at least not when compared to how it turned out. My first encounters with Large Language Models for personal use were more playful than purposeful, just messing around after discussing AI with my team. I had also been doing a good amount of reading around AI because I could clearly see real-world applications for how it could be used in our work, to augment service provision. I also completed the Oxford University Saïd Business School course on AI, which fed into the professional applications we are working on.  But for personal use, I was slow on the uptake. In ChatGPT terms, this was during the 3.5 model era, and it felt like a novelty, a toy, a game. I would usually dip into it when bored in airport lounges or on trains when travelling.

At first, I was pushing it into strange corners, absurdities, such as asking it to design and commentate on a FIFA World Cup tournament but using translation companies instead of national teams. CQ fluency was the 1974 Netherlands side: brilliant, great to watch, but ultimately beaten in the final by Lionbridge. Then I made it imagine Paul McCartney uploading his consciousness to the cloud and asked it to list his discography for the next 3000 years. I asked it to write songs in the style of songwriters like Leonard Cohen, Bob Dylan, Bryan Ferry, on silly topics, like making an egg sandwich. It obliged, earnestly, and the output was an approximation, silly but not really authentic. Lots of other people were experimenting in the same way.  

Sometimes it provided responses that were cold and eery, like when it listed the likely pandemics of the next half century (without prompting it provided associated death tolls with each one) or extinction-level events. You could ask it about the cosmos, the infinite and the infinitesimal. Something was drawing me in. I was glimpsing sheer majesty. But, for months, I was using it like an improv partner and not in a serious way.

Then I fell out with it.

The Battle of Liechtenstein

I started testing its knowledge of specific localities and would ask it to describe the route, street-by-street of my old walk to and from primary school, or the layout of places that I know well and so on. I asked it to plot a route from Buchs in eastern Switzerland to Schaan train station in Liechtenstein. Anybody who is familiar with the area knows that this is an uncomplicated 30 – 40-minute walk. ChatGPT firmly declared that there was no train station in Schaan or anywhere in Liechtenstein. I argued that I know for a fact there is one, but it stood its ground. It would not budge. So, I switched to German: “Gibt es einen Bahnhof in Schaan, Liechtenstein?” Suddenly, yes, there it was, complete with a three-paragraph history of it. The language switch flipped something, contradicted itself and it lost credibility to me.

So, I stepped back for a while, although I did continue using it to build upon a business idea.

A Bit About Me

I am a polyglot, but not the fake YouTube kind. No filmed performances in Brooklyn markets with awkward, and frankly ambushed, strangers, no showy set phrases that give the illusion of fluency. It’s just been decades of quiet, serious language learning. A life’s dedication.

I don’t like stating how many languages I have learnt over the years. There is a point where it becomes…somewhere between irritating and embarrassing. It betrays a tragically misspent youth. But, yes, I suppose I am one of those language savant types. There is a point where you realise that you cannot keep them all up. If you have a full adult life – family, friends, work (hopefully in that order!), you can barely keep up two or three languages. I made peace with that years ago. Life matters more than sitting in the dark going over verb tables in Romanian or noun classes in Kirundi. So, for many years I have been content with keeping 2 or 3 sharp and leaving the rest dormant. Then, you can reanimate these ones when you need to very quickly.

A lot of apps sell the fantasy of fast fluency, with promises such as ‘Get fluent in Spanish in 60 days with me, your AI tutor’. To a large extent, it is a con. They only ever take you to a barely functional level. Developing competence in a language is about discipline and cultural humility. It is about awe. You make a promise, an emotional commitment to a whole universe: the speakers, their history, their values, their wisdom and their culture.

The true linguist learns about this awe very early on.

Many Everests to Climb

When you start a language, you need to come to terms with a simple truth: You will never master it. Nobody does. You will never reach the summit. All you get is a good grasp, maybe a high command. You can achieve fluency, but never native speaker ability. Not really, unless the languages were acquired in childhood.

Language is infinite. We are not. Some people find this discouraging, but I find it comforting. Learning a language is like your personal Everest where you may never reach the summit. If you learn many languages, you get to walk part of the way up many Everests. You can take in the view from there. Maybe one of them you can advance further up, maybe three but no more than that. That’s as far as you can go. But look at the view from where you are already!

This attitude has been with me all my life and is infused in my attitude to business, education and lifelong learning. I never went to any of my own graduations. I never treated degrees as pinnacles of achievements. A degree, even a PhD, is just a route to the mountain. It only shows you where you might go. Learning is what happens when you start walking.

If I had five or six more lifetimes, maybe I could learn some of all I want to learn. I curse the span of this very finite life that I occupy. Even so I will keep walking.

Back to AI

My personal use of AI tools eventually became more than silly absurdities. I started using AI tools to help me think, plan, organise. It became my polylingual dictionary and thesaurus, alongside the professional applications that I was experimenting with and are now bearing fruit. Building on these ideas, which were also language-based, but in a health communication context, I began thinking: could I use LLMs to build the kinds of language courses I have always wanted but never found? Not for profit, not for others, just for me: to feed the learner in me, the mountain range I still like to go wandering in, whenever I can.

AI, LLMs and everything they bring are true gifts to humanity. What started as absurdity and silliness, borne out of scepticism, turned into empowerment. From strange jokes to structured tools and curiosity to creation.

And that is the story in the next article.

 Thank you for reading,

Mark Gibson

Leeds, United Kingdom, Easter 2025