‘Einfach machen’ is what I often tell my colleagues. ‘Just do it’, a certain sports brand would translate that motto. Not that we carelessly walk into the next big data security clusterfuck at our company. To the contrary: we have educated and informed our colleagues about how different AI technologies work, what the pros and cons and risks are. We have shown them how to use those tools. And we have encouraged them to play, test, experiment, to see what works for them and what does not. Because you have to find out for yourself to get a better feeling for what AI can do and what it can not do.
There are plenty of companies (and employees) that prefer to talk endlessly about precautions, risk assessments, procedures, data security protocols and more. Without trying and testing these technologies themselves. I often call it moral superiority: the outspoken opinions of those that might be right in theory, but have absolutely no clue about how this tech could work in practice.
I have written an entire book about strict regulation of the tech sector. There is definitely reason for concern, but that does not warrant a wholesale rejection of AI. To the contrary: I believe in innovation, in progress. I also believe that technology is not the solution to all of society’s problems. But it can definitely bring progress.
My start was a slow one as well. But over the past 2.5 years I have integrated AI technologies more and more into my work life. Some things have not worked very well (maintaining my to-do lists in an unsexy Excel sheet turns out to be more pragmatic than doing it with ChatGPT), but there have also been many what-the-fuck moments. Actually, there have been more what-the-fuck moments in the past two years than in the two decades before that. And I am not one to get carried away quickly.
The much-appreciated professor Ethan Mollick, with whom I would never compare myself, has been running many experiments as well. The funniest one is the one where he tries to create an image of an otter on a plane, reading a book. A scene so bizarre one would expect it to be impossible to imagine for AI, as it would have no examples of that in its training data.
And sure, the first attempt was rather laughable and disappointing. But only 30 months later the results are anything but baffling. Mollick can even generate videos now with stunning accuracy and resolution, as if the otter was really flying in the plane. Useful example? Maybe not. But it shows the progress of AI’s capabilities.
Without realizing it, I have been running two similar benchmark experiments as well. The first is asking AI (usually Large Language Models like ChatGPT and Claude) to draw a map of all the first league football clubs from Buenos Aires province, mark their stadiums, the names of their grounds and their capacity. Basically an infographic.
A human being would take quite some time to produce it, because it takes several steps. The first one, researching the clubs and their stadiums, turned out to be easy for ChatGPT. But putting them on a map that I could actually print was and still is, usually a step too far. Gemini managed to do it a couple of months ago, and it looked beautiful… but the location pins were anything but correct. Claude has come close, but the imaging capabilities of my favorite AI tool are still very limited so the maps are a disaster. And ChatGPT is good in integrating online maps… but these only perform when online. Print it in a PDF? Nope, not really convincing. So this is still a hurdle to take.
The second one is a tracker for new concert announcements in Berlin. I have been going to concerts for many decades, but in a huge city like Berlin so much is going on that there is always a FOMO (Fear Of Missing Out). So I have tried to subscribe to newsletters, to let AI subscribe to newsletters, and much more to get automated updates about newly announced concerts.
Until two months ago all of my attempts failed. Claude even told me it was not the right tool for the task. It pointed me to Distill Web Monitor, a tool that works decently but is still a lot of work. So the relief was huge when I started to use Claude Code more and even let it programme applications for me. Over the course of several evenings it managed to come up with an application that I run weekly, across around 25 websites of concert venues. It registers the newly added ones. And serves everything up on a website that initially only ran on my computer. But mission accomplished!
It made me wonder what else was possible. AI thrives on huge datasets. I have created these in the last two decades: the compulsive side of my character kept Excel sheets of all my jogging runs since 2012, my concerts since 1989, or all the vinyl albums in my collection, for example. The creative side of my character wrote six books about travel, football and technology. Built websites about festivals and travel and Berlin and so much more. This was a goldmine waiting to be explored.
So in the past couple of weeks I have tried to come up with ideas to present all that data in a new and hopefully useful and refreshing way through AI. It has resulted in my own playground, an alternative AI version of my personal website. My own otter on a plane.
There are six areas (or rooms), each revolving around one theme. From my work life to travel, tech, football and running, I have tried to come up with examples what one can do with all that data. There are quizzes, animated videos, there is scrollytelling (a short explainer about my AI book). There are graphics and timelines and huge numbers and so much more. And it’s only the start: when time permits, I might try to push the envelope with more interactive and/or creative formats. Why not?
Why all the effort? Because it has taught me so much of what is possible already. How easy it is to have your own applications and websites built by AI. Not constrained by the demands of work, it has given me an even better perspective on the future that is already happening. I have noticed the stupid small mistakes AI still makes. I have been genuinely impressed by the speed and quality of most of the work, actually. And we are only scratching the surface.
So come and have a look. Let me know what you like and don’t like, whether you have new ideas and wishes and dreams. And maybe, just as happened with me after reading Mollick, it inspires you to start trying for yourself as well!
