feat: was ai radio

1 files changed, 50 insertions, 0 deletions

diff --git a/content/blog/2024/was-ai-radio.md b/content/blog/2024/was-ai-radio.md
new file mode 100644
index 0000000..ad2c88d
--- /dev/null
+++ b/content/blog/2024/was-ai-radio.md
@@ -0,0 +1,50 @@
++++
+title = "AI radio was straight out of a nightmare"
+author = ["Michał Sapka"]
+date = 2024-10-31T22:09:00+01:00
+categories = ["blog"]
+draft = false
+weight = 2001
+image_dir = "blog/images"
+image_max_width = 600
+Abstract = "OFF Radio is no longer AI Radio"
+Listening = "[podcast] 2.5 Admins 219: Spooky Stories"
+Listening_Url = "https://2.5admins.com/2-5-admins-219/"
++++
+
+We, at Crys Jurnal are happy to report... no, I'm not doing such scary Halloween.
+
+A few days ago, I wrote [about AI radio](/blog/2024/ai-radio/).
+Well, we can now talk about it in past tense as they gave up.
+The response was so negative, that they are no longer doing it.
+This is the happy part.
+
+But they are not the only ones doing such "experiments".
+LLMs may have proven themselves to be unreliable doing anything, but this won't stop evil people from using it everywhere.
+We can only oppose.
+
+In a completely unrelated story...
+
+Recently, at my day job I was tasked with converting some Scala code to Ruby.
+I tried to do it manually, to actually understand what the hell am I committing.
+But at two occasions, I gave up and asked Chat GPT to rewrite a method 1:1.
+It did it poorly, but after some back and forth accompanied by cursing, it worked.
+I've been told, that even for Scala devs the code was convoluted.
+But at a result, the Ruby code was convoluted.
+It looked like they hired a Java guy to write Ruby.
+It works, technically it's correct... but it's not Ruby.
+Therefore, I added a comment
+
+> Warning. This method was converted from Scala code by LLM
+
+I may have played with the devil, but this allowed me to feel better about it.
+I even had to explain myself during patch review.
+One thing I didn't do, was to normalize its usage.
+But, in the end, I noticed that it didn't saved me any time.
+I still needed to refactor it, understand what the original code did, and test it.
+All it did was adding uncertainty.
+
+So, in my book, one of things LLMs can't do reliably is helping in coding.
+This was the first time I've tried to use Altman to help me in work, and it was a failure.
+Just like AI Radio.
+This is the sad part.