Thoughts on "My AI Skeptic Friends Are All Nuts"
đź”— My AI Skeptic Friends Are All Nuts
I originally authored a short, title-less post with a few notes. After posting, I thought of a few more observations. I am now migrating those thoughts to a larger post.
Extraordinarily talented people are doing work that LLMs already do better, out of spite.
As Merlin Mann is fond of saying, “Everyone has their reasons." I do not currently use LLMs. The main reason is that I haven’t found a good use case for them in my day-to-day work. I haven’t even found a good use case for them in my hobby projects. That might change someday.
I have a difficult time with the thought of giving up control or giving up the chance to learn, all in the name of expediting something that I actually enjoy doing. If you find value in these tools, then I am happy for you; I currently don’t see much value in offloading things that I enjoy doing. Additionally, there is some evidence that use of these tools actually reduces your cognitive ability over time. This seems like a double whammy: LLMs do the work, hampering your ability to learn and in cases where you actually are potentially learning, you’re not retaining it. I sometimes joke that the United States is hurdling towards Idiocracy; with tools like this, we actually might be accelerating that trajectory.
People complain about LLM-generated code being “probabilistic”. No it isn’t. It’s code.
Fair point. It’s code that the model was trained on from humans who wrote it on Stack Overflow. There are lots of derivative ideas out there, where we take an existing idea and slightly alter it. People will post ideas or share code. They will ask questions and other kind developers will try to give them answers with code samples. This is a big reason that LLMs can “write code”; they’re simply regurgitating what they’ve been trained on. If you’re writing a simple CRUD web application in a common language, these models have been trained on many of these examples. I certainly hope that they can efficiently spit this out.
I have a basic wood shop in my basement. I could get a lot of satisfaction from building a table. And, if that table is a workbench or a grill table, sure, I’ll build it. But if I need, like, a table? For people to sit at? In my office? I buy a fucking table.
Agreed. If there is software that you do not care about or you’re working on a project where you have no passion towards it, then this point of view makes sense. Occasionally there are things that do need to be expedited or you just don’t care about the craft. That’s okay. I just think you shouldn’t judge others who clearly do care about the craft.
“I’m sipping rocket fuel right now,” a friend tells me.
Last note: if this friend was saying this unironically, I would find different friends.