Over-reliance on LLMs
With the introduction of LLMs, I find myself reaching for these tools more and more often. It’s just much more convenient to ask an LLM for the answer to a problem than to spend the time working it out myself, especially if it’s something that I don’t really care about in the first place.
That was my initial opinion, that LLMs are a good thing because they can do all the mundane stuff I don’t want to do, and as a result, I would get to spend more time doing the things I want to do, doing the things that actually matter. And in a way that is true. I often use LLMs to bootstrap frontend work, letting it write boilerplate for me while I just “fine tune” the results by hand. This has drastically sped up my workflow because I don’t need to be fighting with css or trying to figure out how to use certain libraries; I pretty much just need to make sure that the LLM isn’t outputting garbage and I can do whatever’s left by hand.
While I do read and understand all the code, part of me can’t help but wonder if I would have actually come up with that solution, or whether I was even capable of doing it. This was never a problem in the pre-LLM days since you wrote all your code, so obviously you could come up with the solution. This really messes with my head because sometimes I’ll come up with a solution, and the LLM will propose some improvements or alternative/better ways of doing it, and I wonder to myself, if I had spent more time thinking about the problem, would I have landed on a similar solution? Since I’m so fast to just ask an LLM these days, I don’t spend enough time thinking about the problem, so I don’t know the answer to that question. It’s really starting to shake my confidence as an engineer, and I’m starting to second-guess my own capabilities.
The general advice for getting better at something, is to just do whatever you want to get better at a lot. It feels like my usage of LLMs is robbing me of the experience I need in order to improve. If you don’t do something for long enough, you’ll start to forget how to do it. I genuinely wonder what the long term effects of LLMs will be on society. At least to me, the LLM seems to be a super-shortcut for getting answers fast.
The thing is though, you don’t learn as much taking a shortcut compared to taking the hard route. Adversity builds character, suffering and pain are good teachers. Mistakes evoke unpleasant emotions that I find much more memorable than pleasant ones. I’m more likely to remember how terrible that one mistake felt, or how frustrating it was to solve a particular problem, so when faced with a similar issue next time, it’ll easier to recall and draw on my past experience. The LLM robs me of these negative emotions (though they do introduce new ones when they don’t work), and thus robs me of the experience of solving the problem.
I feel like LLMs are robbing me of knowledge, of experience, of ability, of confidence. I honestly enjoy doing things the hard way more for the learning experience. The joy in coding is in solving the problem, and I think I’m starting to lose the joy. I think it might be time to put down the prompts and quit cold turkey.