AI Usage and This Blog

Inspired by an idea I think was first publicized by Damola Morenikeji, this page is about my usage of (generative) AI on this blog.

The main idea is to proactively provide transparency and trust in a world where both are increasingly harder to find. While the problems that came with the arrival of mainstream AI tools are fairly obvious, I consider myself a techno-optimist. The tricky part is finding the right balance and putting up the regulatory guardrails at those spots that allow for an even spread of the advantages over the whole of our societies. This usually takes a long time while the technological advancements increasingly don’t, making it especially challenging.

So, more than ever, it’s our responsibility to be part of this discussion.

How Does teesche.com Make Use of AI?

For many years now, I have established two main sections on the website. One is blog posts, the other book reviews, if you will.

My blog posts are first and foremost individual reports about experiences I’ve had at certain races, with new training plans, or with athletic experiments. From the ground up, it won’t make sense to have an AI write it, because the experiences happen in my head, and my head only.

I do, however, use generative AI on my blog posts for help regarding

• finding improvements for building certain sentences in the few cases where I’m not happy with what I came up with
• selecting good synonyms for words I tend to overuse
• coming up with short descriptions of the finished posts fitting for social media
• fine-tuning the headlines so they are short but contain the important bits.

If I would have to attach a percentage to it, I’m quite sure that 99% of the words in any blog post here are written by myself.

The book reviews are a little different, because usually the books I read have been read by huge numbers of others and quite a few reviews about them are already in circulation. The contents are widely available as opposed to the original thoughts that make my blog posts. You might have noticed that my concept in the book reviews I do has changed over the past few years – mostly because of the advent of powerful AI. I used to include short bullet points about each book that summarized its contents in three sentences.

But now, since this is a very simple prompt for anyone to ask any large language model about almost any book, I have moved away from that and started putting the biggest emphasis on my own personal views of the books.

I think that’s more beneficial to readers. While I have played around with the artificially created summaries, I now use generative AI for similar reasons as I do for my blog posts:

• improving one or two sentences when I’m not happy with their structures
• looking up occasional words to use which are more precise than the ones I came up with on my own
• rarely asking about bits of the stories or the thought processes of the authors I didn’t understand in order to clarify my own thoughts.

I am very happy to be able to use these tools for the reasons mentioned and I wouldn’t want to use AI for anything other than that. It wouldn’t make sense. I could probably ask it to write a blog post in my style about a race I didn’t do and it would come up with plausible results, but what good would that be in the end? It’s the personal experience that counts, so it’s easy for me to draw the line here.

How Should AI Make Use of teesche.com?

The same question, just turned on its head. It’s not the original purpose of the /ai page idea, but I still think it’s interesting to think about. One of the challenges with AI right now is that it freely uses publicly available data to train on, largely without regard for the intellectual property rights of its creators.

Me being the author of quite a bit of publicly available knowledge already makes me a victim of that. Only that I don’t see it that way – but that’s a highly individual point of view. Let me explain:

My formative years happened when the internet was just about to gain traction. It was all about community and the benefit to society. That has largely changed, although lots of its elements still remain: Think about the open source software that is maintained for free by thousands of great developers around the world and which powers most web servers. Or things like Wikipedia and WordPress – where people create useful stuff for free for the benefit of all. In retrospect it’s an odd thing that this has happened at all, now looking at the huge companies who try to own everything and put us into their walled gardens, charging for every single thing just to maximize shareholder value.

I am still clinging on to that notion from the early beginnings. The more knowledge people share with others the better. Of course, I still respect people’s choice to charge money for their creations, it’s just that I choose not to. I am happy to contribute my published blog posts to the base of general knowledge, and I’m fine with AI companies charging people for presenting my content in a useful way to them. Yes, they didn’t come up with it, but they make it available to people in a very helpful way.

And just as I benefit from that when I ask ChatGPT about the difficulty level of a certain mountain race, I’m happy that it answers the questions others ask it by using the knowledge from my own race reports.

Unnecessary Bonus Question: How Do I Personally Use AI?

Apart from what I’ve already explained about how I use AI for help during writing, I am a normal personal user of it, too. I am very interested in the capabilities of several different companies, but Claude.ai and the classic ChatGPT are what have emerged for me to prefer on a daily basis. About 90% of my former Google use is now ChatGPT because the things I’d like to know and learn about are most often much more complex than a few (“optimized”) search results can answer. It has been a huge step up for me and I would not like to go back.

Claude I like to use for writing code. I did lots of work on my side project speechbox.chat using Claude, but ChatGPT was also better at some of the specific questions I had. For the countless other little websites I own and maintain, such as bremensanktpauli.de, I also ask specific questions, such as “please write php code which uses imagemagick in order to scan all images in folder x and resize them to thumbnails which don’t exceed 600x600 pixels and save them into a new subfolder” – What used to take me an hour to do, just takes a few minutes now. I love it.

I never use AI to write email or messages to other people (what’s the point).

Last updated on 7th of July.

I’ll update this page should my behavior change at some point in the future.