At the end of the day technology is going to advance, and the rational thing to do is to figure out how to use it effectively. Yes, a lot of technology gets abused all the time, our society as a whole is incredibly wasteful. But I see technological progress as a net positive, if anything I think the problem is with our social structures and broken incentives. And that’s what we should focus on fixing.
For me, these tools have unarguably save a ton of time and frustration every single day. For example, I had to work on a Js project recently for work. I haven’t touched Js seriously in at least a decade and I’m not familiar with the ecosystem, libraries, language quirks, and so on. If I had to figure all of that out from scratch previously, I simply would not have been able to take on this project. LLM completely papered over all that for me. I know how to structure programs, I can read Js just fine, but I didn’t have to spend the time searching and internalizing all these little details of how to run tests, which npm modules I’d need to use, what React lifecycle hooks I’d need, etc. It made the project far more enjoyable to work on, and I was able to deliver it as fast as using languages I’m intimately familiar with.
The thing is that I did have to spend the time to actually use the tool effectively, to develop intuition for tasks it can do well and those it can’t. How to get it to write code in a way I can understand and review effectively, how to see when it’s not doing what I want and correct that. Just like any tool, you have to spend the time to actually learn it to get value out of it. If you start with the premise that you dislike the idea of the tool, then it’s guaranteed that you’re not going to have a good time using it. But it’s a mistake to extrapolate that other people aren’t getting actual value out of it based on that.
Meanwhile, the whole context of this discussion is running local models which are tools that are available to the common person, and do not result in any capture of labor that I can see. You could make this argument with using proprietary models that you rent from a vendor, but it simply does not hold with ones you run locally.
At the end of the day technology is going to advance, and the rational thing to do is to figure out how to use it effectively. Yes, a lot of technology gets abused all the time, our society as a whole is incredibly wasteful. But I see technological progress as a net positive, if anything I think the problem is with our social structures and broken incentives. And that’s what we should focus on fixing.
For me, these tools have unarguably save a ton of time and frustration every single day. For example, I had to work on a Js project recently for work. I haven’t touched Js seriously in at least a decade and I’m not familiar with the ecosystem, libraries, language quirks, and so on. If I had to figure all of that out from scratch previously, I simply would not have been able to take on this project. LLM completely papered over all that for me. I know how to structure programs, I can read Js just fine, but I didn’t have to spend the time searching and internalizing all these little details of how to run tests, which npm modules I’d need to use, what React lifecycle hooks I’d need, etc. It made the project far more enjoyable to work on, and I was able to deliver it as fast as using languages I’m intimately familiar with.
The thing is that I did have to spend the time to actually use the tool effectively, to develop intuition for tasks it can do well and those it can’t. How to get it to write code in a way I can understand and review effectively, how to see when it’s not doing what I want and correct that. Just like any tool, you have to spend the time to actually learn it to get value out of it. If you start with the premise that you dislike the idea of the tool, then it’s guaranteed that you’re not going to have a good time using it. But it’s a mistake to extrapolate that other people aren’t getting actual value out of it based on that.
Meanwhile, the whole context of this discussion is running local models which are tools that are available to the common person, and do not result in any capture of labor that I can see. You could make this argument with using proprietary models that you rent from a vendor, but it simply does not hold with ones you run locally.