The Sudden Electric Servants

It’s a common story:

• Person wonders if products suit needs >

• Tests said products >

• Finds products have potential, but fail for the specific needs >

• Wonders why no one is making something that does >

• Then starts to realise he might be the one to make it.

That’s kind of where I’m at with using AI apps on voice. In theory, they could be an amazing aid for a professional driver. Anything with an auditory and verbal interface is interesting for truck drivers.

So far I’ve tried ChatGPT, Grok, Gemini, Claude and they all fail in different ways, at least for me. They might be excellent for other users, but for engaging an AI while driving all day they are pretty hopeless.

Yet still… there’s that potential…

I’ve been recording my adventures in bot-whispering - occasionally shouting - with daily video and audio evidence, around 120 hours recorded so far. When it became clear that the potential was being unfulfilled, I started to wonder if these recordings might be of interest to developers. I’m still open to that, but I’m also increasingly of the mind that I should create such an app myself.

“But you’re a truck driver, you can’t code!” comes the politely unsung chorus. Actually, I few years ago I did learn some Python, along with some other bits and pieces, so I have a fairly good understanding of the concepts. But that was just a slight move beyond complete ignorance. I didn’t continue with the learning when I realised I would always be coming from behind, if I used it to try and get work. And ironically it seems I would’ve been a prime candidate to get run over by the AI revolution in coding. There’s further irony in the fact that when I went back on the road in 2018, it was truck driving that was being spoken of as the first job to fall to AI.

And it’s that AI revolution that could turn me into an app creator yet. In fact it already has.

At the beginning of last week I hadn’t even heard of Kotlin. And I’ve now created two apps using it (if it’s not obvious: it’s a programming language, what is now used in place of Java for Android apps). That was in just a few hours, on my days off. It included a fully functional data logger, which works perfectly, and comes complete with a totally intuitive UI.

Of course, when I say I created it, I mean I cut and paste the code from ChatGPT, and let it walk me through the entire process.

But hey, they are all my ideas, and I don’t have to give up the day job.

And that - my day job - that is my strength in all this. I know, in great detail, precisely what a working driver actually needs from an AI voice app. I’ve got my own subjective observations, and now reams of objective data to back it up. That’s what the data-logger is for, to add to that stack. I don’t need to do expensive market research into what truck drivers want, that’s the thing I know best of all.

At first I’ll just make something for myself. After this extensive testing of AI apps, I now have a long list of what needs to change, and what aspects need to be added. What is more, I now also have a good idea of how this could be designed. And how do I know? Because I talk it through with the AI voice apps! When I came back to driving in 2018, I realised that it was an opportunity to learn, at first mostly through audiobooks, so passive learning. I never imagined that such deeply interactive learning would become available.

And what am I learning about? Well, over this last week I’ve been learning about Large Language Models themselves, how they work under the hood, how their developers steer them, how they interact with apps, and how those apps can be built.

So now I can have a go at building it. Seeing as I suddenly have these electronic servants. Yep, that’s how I see AI apps: as if I suddenly have access to a 50k a year Computer Science professor, a 60k a year business studies professor, and a team of 100k a year tech guys. With this in mind, it would seem a total waste to not build something.