Post

I had, up until recently, said that the tools were still a net negative, and I do still believe this is true for anything other than small codebases, bug fixing, transpiling, proof reading, and translating (maybe a few other domains). But it is hard to ignore its utility in those spaces.

I think in many ways the biggest problem is the temptation to use it, like you said. It's just easy to offload your brain to it, which is really the worst thing you could do, because it's not *that* good. I've been saying it's the leaded gasoline of our generation.

i'm not sure it's that bad... leaded gasoline killed untold millions of people. i think many of the issues with AI are actually easily fixable

i feel like your arguments for using these tools in your work will also apply to other uses that are currently less socially allowed and easier to detect, say for images, video and audio assets. i guess my question is,

if in the future those uses become more permissible for a larger audience or harder to detect, why wouldn't you use those tools if you or someone else argues that the creation of these assets is "uncreative, grueling labor that is exactly solved by these tools"?

images, video, and audio are all very different from code. code isn't representative. i also think there's a difference between using models to generate a painting and using them to generate misleading real-looking videos of people.

"despite many claims, programming is absolutely full of uncreative, grueling labor that is exactly solved by these tools. i believe using them is the humane thing to do." it's funny how so much of programming history is just people trying to find different ways of avoiding that grueling labor

personally, i think a lot of the issues people have with llms come down to the people at top having 0 fucking idea what problems people want to solve in a field and marketing being incredibly tone death as well as the whole moral issues but ive seen enough people dismiss moral issues for convenience

copilot was popular within programming not because llms are more accepted because of OSS or w/e its because it was billed as a way to reduce tedium and boilerplate modern programming llms are too focused on being full replacements instead of tools which is why there's a bigger split (in my pov)

it makes me feel crazy when someone is like "if you ever have to write boilerplate, you are programming wrong". like dude...

i think the problem is that a lot of the yapping you see online about this kind of thing is from hobbyists who can't even conceive of the types of problems you have as a professional

...were you gonna say Fridgle is vibe coded anywhere other than a blog post orrrrrrrrrrrrrr

its pretty obvious just from looking at it no

the fantasy world you wish for at the start of your post, under which these tools would be ethical to use, is not the one we live in. this sucks!

i agree. but we also live in a world now where *not* using them in software is unethical too. when LLMs can find vulnerabilities in your software that humans can't, humans alone can't protect against those exploits. i don't really like this future either

🫂 i'll defer to the web experts on this one, i only make locally run toys. i hope you reconsider giving further money to the mega corp now that the experiment is concluded.