summaryrefslogtreecommitdiff
path: root/memos/WM-072.txt
diff options
context:
space:
mode:
authornetop://ウィビ <paul@webb.page>2026-04-11 14:24:49 -0700
committernetop://ウィビ <paul@webb.page>2026-04-11 14:24:49 -0700
commit8c34d810af95fae0ef846f54370a8c88bfab7123 (patch)
tree436beaf30f7b2b3f15741dd54a37e313964d1f7d /memos/WM-072.txt
initial commitHEADprimary
Diffstat (limited to 'memos/WM-072.txt')
-rw-r--r--memos/WM-072.txt93
1 files changed, 93 insertions, 0 deletions
diff --git a/memos/WM-072.txt b/memos/WM-072.txt
new file mode 100644
index 0000000..6a6768f
--- /dev/null
+++ b/memos/WM-072.txt
@@ -0,0 +1,93 @@
+
+
+
+
+
+
+
+Document: WM-072 P. Webb
+Category: LLM 2025.07.07
+
+ In Defense of "AI"
+
+Abstract
+
+ The hate train looks fun but maybe relax
+
+Body
+
+ Over the weekend I saw a post on Mastodon listing a few reasons why
+ LLMs are terrible (stolen data, power hungry, speculation bubble,
+ ruining your brain). I won't refute these points because everything
+ can be true and untrue at the same time, just depends on context; for
+ the record, replace "LLMs" with "people" and the solution isn't to
+ get rid of people.
+
+ I've been using LLMs (I dare not call it "AI" because…that's not
+ intelligence IMO so why scaremonger?) for maybe a year or two and
+ loving it for the most part. I started with ChatGPT, then spending
+ money on it, not liking the results, to trialing Anthropic's Claude,
+ to subscribing to their service ever since. For background, I am a
+ self‑taught design engineer. I got my start in the recession of 2008.
+ I couldn't find work so spent a lot of time on the internet imitating
+ the cool stuff my friends were doing on deviantART. Anyhoo,
+ fast‑forward a decade and I'm utilizing StackOverflow for answers to
+ questions I don't even know how to formulate sometimes…and getting
+ absolutely shit on for it. There was a time period after I was fired
+ this one time[^1] that every question I posted to the StackExchange
+ network got downvoted with no comments, and this happened
+ for months).
+
+ You know who doesn't give a shit about your dumbass questions? LLMs.
+
+ The best of humanity is great. We are the culmination of hopes,
+ dreams, aspirations, and the wide range of emotions that flourish
+ from that. On the flip side, when can get nasty and downright evil if
+ motivated enough. I don't need to suffer through YOUR bad day because
+ you happen upon my naive request while in a horrible mood and you
+ can't help but put me down, or whatever the case may be.
+
+ This is pretty much my argument for why LLMs aren't totally bad,
+ haha! There's not much space for nuance on the internet these days so
+ your opinion isn't swayed if you think LLMs are a scourge upon the
+ Earth. The same was said for television in defense of newspapers and
+ radio. It'll be fine. Railing against "AI" is fine but to pretend
+ that it isn't useful is a crock of shit. Speaking for myself, I am
+ not an expert so having an approximation of one at my disposal at any
+ time of day or night is an indispensable tool I happily utilize to
+ figure things out or get research about things I've wondered about.
+
+ Case in point, I'm thinking about starting a magazine. Claude
+ researched for about 5 minutes while I did something else. Could I
+ have done this myself? Of course. Would I have a comprehensive report
+ in 5 minutes or less? In this (search) economy? With Google,
+ absolutely not. With Kagi, possibly. I could also turn screws
+ manually or use a power tool. It doesn't matter.
+
+ For another recent project, I got Claude to scaffold me a video
+ transcoder built on ffmpeg and it didn't work, no matter how many
+ times I badgered it to fix its issues and stop adding mystery
+ functions. The transcoder works now of course, but the logic was
+ fixed by me, a human. An enthusistic junior developer could make the
+ same mistakes Claude makes but understands, "Hey, this isn't
+ working," doesn't mean tear down the entire project, rename some
+ variables, and call it a day.
+
+ I resent the phrase, "AI is stealing jobs," because no, they aren't.
+ Middle managers and CEOs who read articles light on substance about
+ the marvels of AI are firing people, realizing AI is not a
+ replacement for people, and hiring people back.
+
+ You still need to **understand how to code** to make LLMs work for
+ you. These tools are NOT a drop‑in replacement for people, nor will
+ they be for quite some time. As a replacement for elitist cis‑White
+ male dominated Q&A spaces, my Black ass will take an LLM any day.
+
+ 🕸️
+
+References
+
+ [^1] In 2015, my wife (then girlfriend) and I experienced a
+ miscarriage and I didn't know I should take time off instead of
+ being depressed at work and after given a throwaway project to
+ work on (a website for the CEO's father), I was fired.