As we are approaching the end of No Sloptober - the challenge of going through October without using any LLM based tools - I thought I would reflect on my experience thus far. These tools have somehow crept into my life and they seem almost impossible to avoid, being pushed by workplaces, search engines and infuriating customer service chat bots, but I’ve spent the last four weeks doing my best avoiding them and I’m ready to share my thoughts.
Just Ask Chat
Prior to taking up this challenge I’d find myself in a few situations where I’d resort to using a coding agent or a chat bot, and I hate to admit that the most frequent scenario for ‘asking chat’ was me being lazy. Of course there would be other scenarios where I’d pick it up, like using it for writing boilerplate code, as an advanced search engine, traversing documentation or just playing around with a genuinely cool technology, to asses what it’s capable of - but most of the time it would be due to my laziness. It’s not because I want to be lazy, but I suddenly found myself with access to a tool that makes me able to solve complicated problems without exerting myself. That mental capacity could of course be reallocated to verify the output of the LLM, but when it gives me something that seems to work and looks good at first glance, why bother? That is probably not a great way to use these tools, and I’d agree - but that doesn’t change the fact that this was the way I’d often use them, and anecdotally how a lot of people around me uses them too.
The more I used the tools, the more of a habit it became - maybe even a slight addiction. Ask a question, wait for a response, use the output. Often only little thought had gone into the question. In case it didn’t quite work, I would try again. I’ve spent my entire life learning, but was it all for nought? Why do I need to think, when I have a machine that’s able to think for me. While I am exaggeration a bit, a feeling of existential dread did sometimes sneak in because the models have become really good at emulating intelligence and I’m unable escape the fact that they can produce source code and text much faster than I’ll ever be able to.
Around me it felt like the majority of people were excited and optimistic about this new technology, using it for writing e-mails, code, generating images and slideshows - even settling discussions with a ‘chat agrees with me’. I didn’t want to feel left behind, and this many people can’t be wrong. Right? All the AI positive media coverage occasionally convinced me that being sceptic towards AI is the same as being opposed to progress. That maybe I was just becoming an old contrarian resisting change. I kept using the tools, however I’ve definitely had an internal conflict.
Shelving the chat bot
When the first of October came round, I decided to shelve every tool build upon LLMs for the month, participating in the No Sloptober challenge. I’m on a mission to rely less on the tech giants, so it wasn’t hard to convince myself to take up the challenge. I also wanted to prove to myself, that I was still able to think, without using chat as a crutch when problems were hard or my brain was fatigued.
Generally I wouldn’t say it was hard to abandon the agents and the chat. However on days where I’d experienced a high level of cognitive load, or had tedious tasks at hand, a little voice in the back of my head did whisper: “It would be so easy just getting chat to do this”. The urge to press the easy button was absolutely present within me - I’d become accustomed to offloading boring or mentally taxing tasks. This sometimes left me quite frustrated, knowing I could just throw a few sentences into the machine, and I’d be able to move on to the next thing on the agenda instead of wasting my time and energy.
The recurrence the feeling of frustration did however end up being somewhat of a revelation, it made it very clear to me how reliant I’d become on these tools in my day to day. When this feeling would arise, I’d take a step back, make a cup of coffee or do a bit of exercise to gather the mental energy needed to tackle the problem in front of me - manually.
What Will I do in November?
I’m not denying that LLMs are useful. They most certainly are. I’ve recently been getting into the PlayStation 5 homebrew scene and watched AI give a lot of people the ability to create fun and creative projects that wouldn’t have seen the light of day without it. However I’m also of the opinion that we need to think critically about how we decide to use these tools, and not let our brain wither away by delegating knowledge and understanding to the machines and the companies building them. The companies selling the tools are interested in having us as repeat customers, not teaching us to think for ourselves, and we shouldn’t forget that - they are not our friends. I also have other problems with the AI companies, like how they are stealing information to train their models and profit off of that stolen knowledge. I guess that might be a rant for another day.
I’ve decided going forward that I’ll be very mindful of how and when I use an LLM to assist me with solving problems, and try other avenues before resorting to tools, that might be so ridiculously expensive in the future, that I wont be able to afford them anyway - take a look at recent data center investments, and tell me the prices will stay low.
Instead I’ll try to treat learning new things just like I treat playing video games. Half the fun is dying, learning the mechanics, and finally beating the boss I’ve been stuck on for hours, with my heartbeat sitting all the way up the throat, because I’ve invested so much time into honing my skill, and it has finally paid off. Alright, alright, learning is probably not quite as fun as playing video games, but I definitely remember a time when I was extremely excited about all the things I could do with programming, and I want to restore that feeling. So I’ll expose my brain to resistance, train it like a muscle and abstain from always taking the easiest path - and I encourage you to do the same.