Comment on An earnest question about the AI/LLM hate
blackn1ght@feddit.uk 2 weeks ago
I feel like it’s more the sudden overnight hype about it rather than the technology itself. CEOs all around the world suddenly went “you all must use AI and shoe horn it into our product!”. People are fatigued about constantly hearing about it.
But I think people, especially devs, don’t like big changes (me included), which causes anxiety and then backlash. LLMs have caused quite a big change with the way we go about our day jobs. It’s been such a big change that people are likely worried about what their career will look like in 5 or 10 years.
Personally I find it useful as a pairing buddy, it can generate some of the boilerplate bullshit and help you through problems, which might have taken longer to understand by trawling through various sites.
taladar@sh.itjust.works 2 weeks ago
It is really not a big change to the way we work unless you work in a language that has very low expressiveness like Java or Go and we have been able to generate the boilerplate in those automatically for decades.
The main problem is that it is really not at all useful or produces genuinely beneficial results and yet everyone keeps telling us they do but can not point to a single GitHub PR or similar source as an example for a good piece of code created by AI without heavy manual post-processing. It also completely ignores that reading and fixing other people’s (or worse, AI’s) code is orders of magnitude harder than writing the same code yourself.
blackn1ght@feddit.uk 2 weeks ago
If we include languages like C#, javascript/typescript, python etc then that’s a huge portion of the landscape.
Personally I wouldn’t use it to generate entire features as it will generally produce working, but garbage code, but it’s useful to get boilerplate stuff done or query why something isn’t working as expected. For example, asking it to write tests for a React component, it’ll get about 80-90% of it right, with all the imports, mocks etc, you just need to write the actual assertions yourself (which we should be doing anyway).
I gave Claude a try last week at building some AWS infrastructure in Terraform based off a prompt for a feature set and it was pretty bang on. Obviously it required some tweaks but it saved a tonne of time vs writing it all out manually.