AI Fatigue

Backstory

I started my self-taught web developer path around 4 years ago with The Odin Project, which is an amazing and free resource to learn “How to code”, specifically web dev. I never finished the curriculum, but excluding those who gave up, this seemed like a fairly common path to start working on your own type of projects or even start applying to jobs. Even one of the youtubers I watched, Self Taught Dev, landed a job before finishing. It seems I can’t pull up his channel, but here is a random video by him -The Odin Project - Is it the best way to learn web development for you?

As for myself, according to a chart I found, I started around when tech job postings peaked. Depending on your perspective, I had just either just missed the train or didn’t learn quickly enough to beat a contracting job sector. I sometimes wonder if I currently would be better or worse off if I had joined a coding boot camp the moment I quit Amazon.

ChatGPT Releases Nov 2023

It was pretty cool to use this new tech as it released. It helped me with certain aspects of my first website but even then, perhaps just shy the skill level of an entry level dev, I noticed how often it would be wrong or even give me detrimental advice/code. I always tried to keep it as more of a mentor by critiquing my code, or asking why this or that solution or how this implementation compared to this idea. There are two examples in my head where I had reached a frustration peak and just kept copy pasting code and errors messages into GPT like I was spinning reels at a slot machine.

The first was Ruby on Rails Asset Pipeline (Whats the biggest headache with a rails project?), which I believe I had mixed up importmaps with esbuild. Skipping any technical jargon that I also probably only understand enough to get by, this was a nightmare. Eventually “we” settled into a solution but I understood almost nothing about the Rails asset pipeline at the time.

The second was launching said website on AWS’s (Amazong Web Services) VPS (Virtual Private Server). Basically, I rent a fraction of a computer/server resources to run my website. It’s not as straight forward as using resources like Heroku or Fly.io, but I also thought it would look better on my resume.

So I wanted to use NginX to route traffic but I could not set it up correctly no matter what I tried, even after close to two weeks of troubleshooting (which I did other things as well, but that was the big focus). Toward the end everything seemed like it should work and the settings were correct but I had just become a monkey pulling the slots lever, hoping for a win which never came. I eventually swapped to Kamal which is the default for Rails apps anyway.

Where’s The Rant?

Don’t worry, it’s coming. But this is my blog and I wanted to discuss my background which I think for some people your technical ability matters (Senior, Junior, vibe-coder, etc.). I don’t use AI for anything creative, outside of Canva for removing backgrounds on images or erasing Text/Game UIs for my thumbnails.
Which I assume should be fairly evident that my blog entries and my youtube thumbnails are all 100% human-made garbage (maybe a tad harsh, but I find the line funny).

I still discuss certain things with GPT or Claude (which I only just tried Claude for the first time in May - better for coding but its whatever) when programming, but I’ve been using far less this year. They have their use cases and can speed up certain aspects but you need to know how to use it effectively and recognize if they are giving you a bad or less than ideal answer. Like a different way to “googling” something.

The thing is though, if you are trying to learn then you need to use them far less or only for a top-down perspective. Like when going through some of the later projects in The Odin Project, requirements became more and more vague on purpose so that you were forced to fill in the gaps or hold yourself up to a certain standard. Projects would use skills taught in prior lessons, but also purposely left certain things out of lessons to force you to learn how to google/research how to solve certain problems.

I already didn’t use LLMs like a script kiddie, copy and pasting everything without understanding. But anytime I use it a shortcut, no matter tiny (like coming up with variable names) I’m only cheating myself out of learning. Especially the more sucky or frustrating something that should work but doesn’t is where a lot your learning will come from. Pushing through that pain is like doing a rep at the gym or when meditating noticing your mind has wandered.

Now I don’t think AI can account for all or even a majority of the tech market job contraction, but I would find it hard to imagine it’s zero. It would take some responsibility off my shoulders but even so its still fully on myself to either get a career in tech or not, regardless of a bad market or if things are “fair”. Although it is possible it has delayed getting back to normal (pre Covid era) by who knows how long.

I think AI has just be used as a scapegoat for companies over hiring and just like a pendulum it will swing back. Of course some companies are probably just not doing well and as a result do layoffs. I found it interesting when Primagen talked to Big A - 28:14 about some company layoffs, Intuit (developed QuickBooks) specifically said these layoffs were non-AI related and their stock prices “cratered”.

Rant

Finally at the point of the post dealing with the title!

I hate the AI voice in any youtube video and instantly stop watching (except for Core Dumped), and can be quite annoying with certain voices that are borderline either monotone or quality AI voice. I find myself distracted and focusing on how the voice sounds rather than the content of the video itself. I have more than enough channels to waste my time if needed so I also move on from questionable voices which might not be fair but what can ya do.

I dislike how enamored the older generation tends to be with AI, but I suppose its no different than just slop content which existed pre-AI. But I also hated all that content too and there’s plenty of hate to go around for low effort content, which became even more low effort. At least this is made up by the next generation hating on AI. As since recently with all the booing at college graduation with boomers talking about AI and the future with AI ect.

My father will share YT shorts slop content far too often but I will still watch a small selection of them because he will also watch things I share to him. Does the AI generated videos that accompany stories provide any value? What is the point of these visual aids that weren’t made with purpose, but rather just lever-pulled into exist at the slot machine. Choosing the best of the bunch, if even that much effort was put into making these at all.

Once I get a hint that AI was used in any capacity, with thumbnails as the possible exception (quite dislike it still), I basically question what else was AI used for? When the videos my father links either have AI voice / visuals, I just assume the script would also be AI or just the Wikipedia article brought to life (well life is a strong word, more like a Frankenstein monster).

We live in a society (shocker) and I don’t care how people spend their money, but it is a different beast entirely with the sheer magnitude of money (and man-hours, but in a simplistic way I just view money as an extension of man-hours) that has been spent on AI and more recently their Data centers. Many things could have been built in their stead, such as nuclear reactors!

I’m sure I also don’t even need to go in detail how this has affected GPU, RAM or SSD prices. My gfs system is fine (was pretty decent when it was built in ~2019) but she would’ve upgraded within the last year or so if not for the pricing. I’m just glad I was able to get two 1TB nvme’s (for linux and youtube video editing) before it shit the bed.

Now I don’t know quite enough on this topic but I’ve heard Ed Zitron talk about how a lot of these Data centers aren’t being built at the pace talked about and probably will never get started. Fraud? Enough grass-root pushback at the local level? These companies believe the bubble pops before the centers are finished? Or rather lost hope in making profits with AI so they don’t want to dig the hole even deeper?

Github Copilot Charging By The Token

To take the line from this reddit post, I also feel a little Schadenfreude when hearing about the downfall of companies or even individuals who rely too much on the use of AI. Do they not know how much AI is subsidized? I mean it’s quite hard to know exactly how much, but are they aware that it should be more expensive at all?

I have never paid for AI and I pretty much never will. I could foresee a future where paying a nominal fee allows you to be more productive with selective use of AI, which time equals money (especially if you’re freelance as opposed to a salaried job). But in my current circumstance where I’m learning and attempting to break into the field, my use of AI would become zero if the free tier was taken away.

June first Github Copilot swapped to direct token billing. You pay for the resources that you burn using AI, novel idea that. As seen in that reddit post, companies are burning through their allotted AI costs faster than expected. Uber spent their years budget by April and many others companies are begging to restrict their token spending. Capping how much engineer can use and/or swapping to older and cheaper models.

But then how are companies like OpenAI supposed to make money and make these data centers worth building if people use models that are way cheaper and maybe ~%15 ‘worse’ than the newest model? Building data centers to make tokens more affordable doesn’t sound like a winning business model to recoup all the years or losses thus far. Users can also turn to local models, which to my limited understanding is GPU VRAM and RAM limited with how fast they can run. In simple terms, stronger rig means strong AI but stronger rigs are already quite expensive.

Rant cont.

I’ve been writing this a little too long so I’m going to wrap this up. Perhaps I’ll sum up all the actual AI fatigued/AI hate I experience in its own separate post (nearing 2,000 words so congrats if you made it this far).

I’m a little risk averse and maybe people expect a “rug-pull” with OpenAIs eventually IPO and are just trying to buy in and out quickly? Maybe if I had some amount of FU money my strategy would’ve been to buy at asking price and set sell offs with half at 10% and the other half at 15% to try to make a cool 12.5% easy money. But I think some people are just all in with AI and expect it to go to the moon?

Although I guess I don’t have much sympathy for people who risk and lose way too much money with these IPOs. It just rubs me the wrong way that important and rich people at these companies will get to offload their shares and leave the bag with the people.
Turning the world upside for a few years and making money doing so.

A whole other topic could be all the chatbots for lonely people, but that’s more of a symptom of our society and not because the chatbots are amazing. Maybe some people like having their egos stroked but I despise when the AI tries to glaze me its all so fake, “Now you’re thinking like a developer”. It’s not just X but Y–and here how Z plays a role.

I also hate that anytime I see an em dash or certain sentence structures my brain goes right to AI. Even though em dashes were used before AI (hence why they use it so much, its in their training data), its quite a prospective tell.

End This Please

I’ve rambled long enough. Its quite cathartic to write things done, something I haven’t done too often. Even for the few videos I’ve written a script for I went off script or changed words around all the time when reading, so I began just writing bullet points, like a classroom power point presentention. Yes I was one of the kids who actually talked about the slides and didn’t just read off the slide.

I just got a Lark M2 for my birthday so I’m excited to do the verbal version of this on my youtube channel. I’m still not sure exactly how I will relate my blog posts to my videos, such as:

• Making blog post after making a video about it (so expanding the topic or more of a transcript)
• Writing blog posts which I then translate into a video (either word for word or just as a first draft that I refine)

Either way its nice to get back into creative work. I still haven’t made a youtube video in almost 2 months but it’s better as hobby vs setting deadlines I try to meet, even though it is beneficial to having some sort of schedule as being creative is its own skill.

Enjoy the rest of you day.