A short vent on LLMs
Admittedly, I use language models' services daily to understand and write code. I wouldn't be able to contribute as many features as my other coworkers, and I'd likely be underperforming in the eyes of the corporate. And that would, in turn, put my family's well-being at risk -- I just can't afford that.
Yet, I can't get away from the feeling I'm trading my soul bit by bit, every single day. I can't tell whether the ideas that go into merge requests are mine, or ones of the model.
I feel at the service of the machine, a heartless probability computer, begging it to produce an acceptable outcome.
For me, the whole situation resembles engaging with your own trauma: you feed time and energy into something in, hopefully, exchange to not torment you as much -- but deep inside you know it's a dark well no one can fill.
Perhaps, I could take a brief AI detox for a week or so.