I am a scientific research engineer (bioinformatics), and here no one cares much about covering all the possible code paths.
What we care about is if the code computes "the correct thing", i.e. that it represents the underlying science.
No such guarantee with LLMs. But no such guarantee without LLMs, either (the "code growing above our heads" has happened already, a long time ago). Still, I would say that LLMs are a big net positive for us: they are better at checking such things than we are.
A dirt ramp can help you get a truck up and over a 2 meter wall, but a properly built concrete ramp will work better long term.
Determinism is what CPUs excel at; when they dont, its because of either bugs in the hardware, or in the software. in theory, those can be contained.
but the allusion that theres some third magic is silly.
LLMs are problematic because of the addiction, yes, but more so because they reas stack overflow and dont discard all the hacks, so those will srface. when you tell it "dont make mistakes" you might be pointing it to those underlying comments that say "OP, you forgot to..." or "the cuerent API is.." or "its 2025, we can now do x"
The major point on the nose is the LLM is working inside the branching logic but its not going to follow control flow. it has zero idea how anything will compile beyond a few ifs. And only if you canslice and dice the proper context like a DAG where its no poisoned by irrelevant but badly named methods or constants polluting its scope.
They basically deny everything positive AI brings. If AI cures cancer, they'll say but AI hasn't cured aging yet so it's still useless. If AI solves a math conjecture, they'll say but AI hallucinated that one time so we can't use it. The goal post keeps moving. It's the opposite of r/accelerate.
When I read the sub, I can't help but get a cult-following feeling. They'll twist facts and bend them to their beliefs. It's not much different than people who seriously think the Earth is flat in my opinion.
There are so many more concerns around LLM use that for a lot of people the goals haven’t moved at all and that it remains there is no ethical use of A.I. not the least of which are:
There is no argument for A.I. that doesn’t amount to providing capital access to labor without providing labor a reciprocal access to capital.
The source material was ingested without permission or license of the copyright holder. No residuals are being paid. If your industry relies on crimes to exist it probably shouldn’t.
Hyperscalers are adding data centers to under developed and rural communities dramatically raising the cost of electricity while consuming water and providing few jobs.
A tragedy of the commons as the internet balkanizes into more and more walled gardens specifically to prevent the ingestion of intellectual property.
the context is business and AI jesus, eg, AI will fix all the problems AI has.