AI Skeptic Friends

Saying hallucinations are a solved issue is a crazy take

"every line of code is a liability" I really wish more people, engineers or not, understood this

"Hallucinations are solved" in the same way that fusion is "just an engineering problem now"

AI hype is like saying "We don't need doctors anymore because we can google symptoms".

"nobody cares if the logic board traces are pleasingly routed" is a great software-only take. the electrons care

A lot of people seem to think a "Junior Developer" is someone who rolled out of bed and decided to code. It's actually someone who has a whole lot of experience coding, probably some kind of 4 year STEM degree, but hasn't yet proven themselves in a professional setting yet and may be learning some new tech. If you just went to a bootcamp for a few months to learn how to code never having touched a compiler before, you're not even a Junior Developer.

The AI solved the unit tests by deleting all the unit tests. So productive.

lol, the point of teaching junior engineers is not the intellectual activity of improving code, it's that you're uh... teaching someone

LLM AI is just like someone who only knows the correct answer 70% of the time but answers confidently every single time.

As an industrial engineer who writes safety critical code code for embedded machine controls I have a healthy skepticism of LLMs. The main area where they completely fall over is their inability to come up with any genuinely novel solution that has not been submitted to Stack Overflow already. Just out of curiosity I challenged a few different models to come up with an algorithm I worked on recently for predicting heat transfer in a real world system. All of the models came up with complete nonsense and even when told to use least squares regression they failed to come up with working code that fit the project requirements.