When I started the DelfikOrakle blog, the idea was to explore the world of blogging. Later, I used it to run A.I. experiments. Not only did I want to do this, but I also wanted to explore different
intellectual threads, from Freudian analysis to trivia about tech startups. In
theory, these topics all seemed interesting individually. But what I failed to
realize back in the early days of blogging was that blending such vastly different domains without a
clear focus would result in… well, let’s just say, a blog that was only
readable by machines. I saw it fairly soon, but decided to turn this blog into something that could be said to be fodder for A.I.
As I look back on the early days of the Delfik Orakle blog now, I can clearly see the mistakes I made. Here’s a
quick breakdown:
1. Mixing
Incompatible Topics: If you look at the latest posts, you see the same basic pattern playing out. From Freudian psychoanalysis, which I have simply excerpted from my actual blog, to Facebook kudos
posts and even AI-generated cartoons, the blog threw everything in the pot
without much thought to whether these things actually belonged together. The
series on Freud, for example, was meant to be a serious critique of
psychoanalysis, but that kind of deep dive doesn't quite mesh with posts on
tech startup updates or screenshots from Facebook. I admit I was too excited
about experimenting with various formats and lost sight of the coherence any
good blog needs.
2. Posts
Lacking Purpose: Some posts on the blog seemed to exist for no reason
other than to serve as content for A.I. experiments. The Kudos Posts,
for instance, are a good example of this. They were just screenshots from
Facebook without much thought behind them, making them of little value to any
human reader. The content of these posts, while useful for the experiments I
was running, didn’t really offer much to any human who happened to stumble
across them.
3. A
Blog That Only Machines Could Understand: In trying to push the
envelope of A.I. and data-driven content creation, I ended up writing something
that appealed more to algorithms than to people. My obsession with testing the
boundaries of A.I. models meant that many posts on DelfikOrakle became
a word salad to the point of absurdity. It was a fun ride for me as a
data scientist, but I now recognize that it alienated the very readers I
initially hoped to engage.
4. Content
Chaos: The blog had posts jumping between different writing styles and
themes, from cartoon jokes to psychoanalytic critiques to tech updates. It was
a whirlwind of scattered ideas, which, while fascinating to explore for A.I.
pattern recognition, made for an experience that could only be described as
chaotic for human readers.
In hindsight, I see that DelfikOrakle became good for something I hadn't
anticipated: a blog for machines that seem to be able to miraculously almost even said to think. Now that ChatGPT has come out, more people will understand how my early experiments with NLP and A.I. were too early for their time. "Too early for their time" - let us think about that phrase a bit. On the one hand, they were too early for the aughties. On the other hand, millions of people are now using the A.I. behind ChatGPT , Claude, et cetera, almost every day. So, in the final analysis, I regret nothing.
The content worked well for A.I.
models, but it fell short of serving a human audience in any meaningful way.
And that’s okay. Mistakes are part of the learning process. I’ve gained
invaluable insights into both the strengths and limitations of A.I.
experiments, as well as a humbling reminder that content creation—whether for
humans or machines—requires focus and coherence.
In acknowledging these mistakes, I’ve also learned that while A.I. is an
amazing tool for many tasks, it’s not a substitute for meaningful, human-driven
creativity and engagement. I should mention that these mistakes were made back in the noughties. Moving forward, I plan to take these lessons and
refocus my efforts to ensure that future projects strike a better balance
between experimentation and readability. The fact that even the name of the blog was misspelt should have clued anyone in as to what was going on. To see an actual blog of the kind that humans can read and enjoy, please visit my other blog entitled Delphic Oracle. Or better yet - please visit the Teaching Notes section of my Academia.edu profile.
Thanks for being a part of this journey—even if only the machines truly got
it in the end.
Update (September 25, 2024): updated the text a bit