Well, it seems you just don't like programming

I already tried to write a blog post about AI. An article, discussing AI generated art, still festers in my drafts folder. As someone who enjoys 3D modelling, drawing, animating, someone who believes in art lifting our spirits, I find it pointless and distasteful. Not much to discuss.

Here are some loose thoughts about AI, mostly from software development perspective, but not unique to it.

I see coding as both a skill, which enables me to build things I want to build, and an intellectual exercise. Like any other discipline, it doesn’t have sharp borders. If you’re receptive, if you pay attention, you’ll learn things that apply outside of computer science. Underlying principles behind coding tell a bigger story, about systems, math, logic, or effective communication of ideas. Diving deeper into these ideas can become a source of great, personal satisfaction.

As far as I have been able to observe, the next stages of mastery, are locked until you spend enough time exploring your current level. You have to practice, try things, fail, until you understand the nature of these failures and what actually makes them failures. In programming it’s not simple to understand which decision led us astray. Sometimes you make an innocent decision, which impacts the structure of a big chunk of your codebase. Without practice you won’t be able to identify the cause and the effect, and without understanding the issue, you won’t be able to find the solution and therefore progress.

Yes, I’m asserting that AI is robbing you of the necessary practice, to become better. I believe that, especially if you’re just starting, you’ll quickly fall into trap of copy pasting AI’s code, without much analysis. Why would you spend the time analysing a piece of code? You won’t even feel the need of learning, if you never have to solve a problem yourself.

I guess I have made a mistake, repeating “AI” so many times. When we say “AI” we mean “LLM” - Large Language Model. An optimisation algorithm, which looks for a local maximum of what we’d consider an appropriate response. As such, its output reflects its training data, and thus its quality of output will always reflect the average quality of the input. Or, in other words, it’ll always generate a mediocre code. It’ll always push you towards the well trodden path, further hindering your progress.

More like “mediocre TODO apps generator”!

Some say AI democratizes building, making, creating. However, people who have spent time learning a skill, learn what’s worth making. No one exemplifies the lack of this skill better than the AI “art” crowd. Facebook is drowning in an ocean of AI slop. Every now and then we see someone generating images of real people, in pornographic scenarios. New AI models are celebrated every week, yet the images they generate are always ugly, and either pointless or made to humiliate, misinform or scam people. The same twisted, malicious motivations will result in more malware, more scam websistes, more wicked code.

AI seems to be a precise strike, aimed at everything between us and the end goal. The gap between you and your new project should be decimated, say AI enthusiasts. I’d say that that gap is the space where you can exist, express yourself. It’s a truth that we all know, but rarely internalise - it’s the road that makes us who we are, not the goal. All great coders, artists, creators had a lot of stinkers between their best work. That process of trying, failing, learning, changed them, sharpened their senses, built their understanding, elevated them to new levels of their craft. If you remove the process of making, not only you won’t grow, but you won’t develop your taste, to come up with better ideas. You’ll be stuck in the “TODO apps land”.

It, however, tracks. In the current economical system there’s no space for process, which might benefit the worker. It gives them some space to carve out autonomy. That’s an obstacle, necessary evil between the capitalist and their great idea. Labour creates wealth, but it also consumes a part of it. Wouldn’t it be great if, for a capitalist, if he could cut out all the inefficiencies and costs of labour? Increase of labour force productivity helps us all, but not equally. AI might help you do two times as many tasks as without it. Why are employees expected to be excited for that prospect? I guess that’s because they’ll only have to work half of the work day thanks to AI, right? Or is it because they’ll be paid twice as much as before? Oh, neither?

I don’t know if I have anything to say to you, if you have no interest in coding and you’re here just to pay your bills. Maybe it makes no difference to you, if your role gets reduced to fixing bugs, in LLM generated code. I do not resent this attitude. It’s ok to not be emotionally invested in what puts food on you table. However, this industry is competitive and people who spend their free time learning and practicing have an obvious advantage over others. LLMs will start their gorge at the bottom of the skill pyramid. Your knowledge, strengthen by your experience, will provide some protection, against ignorant managers.

Recently Adam Savage recorded a video, explaining why he builds things. He said:

Something happens in the alchemy of me using my experience and my point of view and my skills and my talent, at the time, to execute an idea that’s banging around in here, and when I do that the thing that comes out of that, it does feel alive to me.

Later he adds:

Personally I don’t think there’s difference between crafting and art, I think they’re just two descriptions of parts of a process. I’d say that one leads to the other, because when you’re good enough at executing on one level, you start to be able to zoom out and see wider. […] At a certain point you stop becoming concerned with all the parameters and you start to be like zooming out on your craft, on your skill, on your execution, and when you are zooming out these thoughts start to occur, of like “wait what if I put a blue stripe there” and there will always be this part of you “I don’t know where that came from”, that’s why I make! I don’t know where that came from, I don’t know what about my point of view knew to do that. That’s another thing is when a piece is really good I often have very little memory of the set of decisions that made it, like, it feels to me like a magic trick.

Adam Savage stays an inspiration. I don’t believe LLMs will ever enable you to experience the feelings Adam describes.

Here are some other people, expressing their thoughts on AI:

To quote a small part of it:

Used improperly, technologies can and do result in the deterioration of cognitive faculties that ought to be preserved. As Bainbridge noted, a key irony of automation is that by mechanising routine tasks and leaving exception-handling to the human user, you deprive the user of the routine opportunities to practice their judgement and strengthen their cognitive musculature, leaving them atrophied and unprepared when the exceptions do arise.