Anachostic

After a lot of pondering and a lot of research and a lot of self-justification, I finally have gotten on board with the AI train.  I admit, I probably did it a little too late, but I also bought my house at the peak of the housing bubble, too, so it’s just what I do, I guess.

Let’s go back to before I had my mixed emotion experience with generative AI.  I would only use Ai sporadically,  like when I had a funny idea and wanted to make a surreal image of the thought in my mind.  That’s pretty much how I viewed AI – just a fun novelty.  I would also later come to use it more and more regularly as a search engine.

That needs a little explanation.  Some people say that AI killed the search engine and some people really bitch about the AI summaries that come with all your search results now.  But, you know what?  Search engines and the broader bullshit gaming of search results by each and every website brought that on themselves.  It is now as it was then for me.  There are two types of query results that I want from a search engine.  Maybe you’re the same way.  I either want an answer or I want a set of data to research.  Old search results suck for the former, and AI results suck for the latter.  If I search for "magical toilet plunger", I don’t want AI telling me all about toilet plungers.  I want to read about cases where a toilet plunger had a magical effect on the situation (or whatever. It’s just an example).  Likewise, if I ask, "Is Metroline part number 123 compatible with my 2010 Mazda MX-5"?  I don’t want a list of sites selling part number 123 where I may or may not be able to find a compatibility chart.  AI can answer that with a yes or no.

And so that’s the extent of my past AI usage.  In earlier posts, I had a revelation with AI music generation – with a lot of conflicting feelings that I had to work through.  I’m not 100% on board that specific train, but I really want to see what it can do.  And while processing those thoughts, I had a really interesting conclusion.  Generative AI could be at the crux of everything about me.  While that sounds like hyperbole, it’s really uncanny how it slots into SO many details of my life.  So let’s run through them.

• Technology – Obviously I’ve been lifelong tech nerd,  I have really fallen behind the times and don’t do much cutting edge stuff like I used to because, well, I thought I’d seen and done it all.  Or at least all that I wanted to do.  AI opens the door to things I’d never considered and while the learning curve is a little steep, keeping your mind active and your curiosity alive has life-long benefits.
• Programming – Beside the fact that AI can write code (which I wouldn’t trust, but I could be a capable proofreader for it doing all the gruntwork, unlike vibe-coders), when you look at the leading tool for AI work, it’s very programmatic.  They say it’s supposed to be easily understandable by non-technical people, but let’s be real.  If you are technical, you’re going to have a much easier time with it.  And the way you learn it is generally a lot of trial and error, exactly like how I learned programming long ago – you change something, it breaks, then you figure out why that caused it to break and you gain knowledge. (side note: for AI processing, you create a "workflow", which is a bunch of steps where the data is routed between them with connections.  It looks and behaves exactly like a modular synthesizer.  Coincidence?)
• Language – This reason came to me as I was learning AI image generation.  The quality of the output is dependent on your mastery of vocabulary.  If you don’t know a lot of words, you’re not going to get unique output.  If you don’t know how to describe something, how can AI know what you have envisioned in your head.  As a long-time writer, this is just another "this is made for me" checkmark.
• Artistry – While I don’t have any graphic design skill, I do have ideas.  And I’m not exactly unartistic, since I’ve written unique music for decades.  Wouldn’t it be fun to have music videos for those songs?  I have some ideas, but I can’t create them – until now.  And the purists that say, "you’re stealing money out of the hands of animators!"  Music is my hobby.  I do it for free for myself and I’m not going to pay someone to create something for me for something I do for free.  That’s like saying you should never work on your car because mechanics exist, or never fix a toilet because plumbers exist.  Get the tools, learn the skills, do it yourself.

That covers the high points without having to be too detailed.  And the result of all that consideration is this idea that building a home AI machine is like when 3D printers became available to consumers instead of being limited to commercial industry.  It’s just instead of physical output, it’s digital output.  Both still create anything you can imagine, a lot of times you’re using someone else’s template, and the quality of results can vary widely.  3D printers got a little hate early on with "Cheap!", "Tacky!", "Who wants that?", but they also were beneficial, like making replacement parts for things that no longer existed.  Maybe someday (like right now, someday), AI will be generating music, video, and pictures of things that no longer exist, with the exact same quality variability as 3d printers.

I talked myself into it.  Initially I decided to buy a used 2nd-tier system (which is still absurdly expensive thanks to AI data centers) with a budget of $3k.  I contacted a couple of sellers on FB and neither responded timely (one still hasn’t).  It got my spidey senses going because of all the reports of people buying high-end GPUs and finding the GPU chip has been harvested from it and sold to China.  So buying used looked like more and more of a liability.  Losing $3k on a "it was working when I sold it to you. You’re the one that harvested the chip!" argument is not interesting to me.  And also, it’s a 2nd tier GPU, which can’t be upgraded without complete replacement.  The upgrade GPU?  $4500 just for the card.

I then convinced myself buying new is safer. So I started sourcing parts.  $4500 for the GPU, hundreds more for RAM, CPU, cooling fans, case, hard drive – there’s no way.  Then I thought, maybe buy pre-built?  A pre-built computer is traditionally seen as the "easy way out", but I’ve cut my hands on so many computer cases in my days building PCs that I feel I’ve put in the time.  And as it turned out, there’s no way I could have come in at that price point. 

Total damage: $5200.  Top-of-the-line GPU, 64GB RAM (with two extra slots to fill later!), 2TB SSD drive, liquid cooled CPU and nice case – I mean it’s a NICE case. Which brings me to today.  I haven’t bought a new PC in decades.  I’ve either just bought upgrade parts or built one myself, but always mid-tier components since I was never a gamer or real power user – just business grade stuff.  So imagine how surprised I was to see an 80 lb box sitting over two feet high being delivered by UPS.  It’s comically huge.

And this is what it looks like next to my old "server", which did mail,web, and database services for years.

So it’s up and running and I’m jumping in with both feet.  But I’m trying to pace myself and learn incrementally without trying to jump straight to the finish.  It’s my old programming habits coming through, like this task is the next obvious step in my life.