Where I'm at with AI
20 points by samuelkarp
20 points by samuelkarp
but the speed with which a lot of industry leaders went from being apparently concerned with the environment to being happy emitting a NYC worth of CO2 in a year is dizzying
This seems like a good example of how capitalism is a race to the bottom. Business leaders are afraid that they have to embrace AI or see their business get eaten by someone else who does. Whatever their feelings are about ruining the environment become secondary to their business imploding, particularly when everyone else is doing it too.
As a professional, I am obligated to use the most effective tools that help me do my job well.
The whole piece would be fine if it wasn’t for this quote. I don’t believe an experienced software developer with plenty of job options is obliged to use any tool. At least not for now.
One should always draw the line on one’s moral compass. The author likes to use AI and should own it, instead of playing victim’s role.
Interesting piece, but I wouldn't agree that LLMs and agentic tools like Claude build on top of them are always and indisputably better and faster than humans at generating code from said specifications - it is skill and domain-dependent. With specs-driven approach, it also takes time to write those specs and then it takes time to review LLMs output - I am kind of on fence there; it is not clear for me that it makes the process as a whole more effective and better overall.
I use LLMs contextually and conditionally; for learning and knowledge discovery - very often, although that definitely has it is limits and shortcoming as well. For programming it depends on the task - often I find myself faster writing code from scratch to be faster; it is a nice help though, but I would not say that it is a revolution; it is just another tool in an experienced programmer tool belt.
Long-term social consequences are the real concern here - will we create less things, because LLMs can? Not only art, but also sharing knowledge and experience? Open-source alternatives most likely will not be able to compete here, since you need to retrain LLMs at least once per year (preferably more often) to be up-to-date and useful and that is prohibitively costly. Or maybe we will be able to come up with some sorts of inventive structure that will make it possible? Interesting times definitely; many questions, fewer clear answers.