In five years time, some CTO will review the mysterious outage or technical debt in their organisation.
They will unearth a mess of poorly written, poorly -documented, barely-functioning code their staff don’t understand.
They will conclude that they did not actually save money by replacing human developers with LLMs.
#AI #LLM #LargeLanguageModels #WebDev #Coding #Tech #Technology @technology
@ajsadauskas @technology lol you don’t need LLMs to end up in that mess … seen it everywhere
NGL, was totally expecting a different last paragraph. 😂
One sentence a paragraph does not make.
@zenkat @technology Totally agree.
But.
It’s a surefire way to get yourself in that mess in rapid time, when you otherwise wouldn’t.
@ajsadauskas @technology AI: do more stupid stuff faster!
@ajsadauskas @technology I agree right up to the end. I think they’ll conclude they need a more powerful LLM.
I dunno, humans are more than capable of doing this already.
@ajsadauskas @technology I’ve been thinking for a while that I truly pity anyone who’s going to have to maintain this AI-generated code.
They’ll outsource to cheaper offshore LLMs.
I think every CTO expects to inhereit technical debt, whether by overworked devs or overautomation
Not five years, but this year itself
@ajsadauskas @technology Here’s some data to back it up too 😊
https://visualstudiomagazine.com/Articles/2024/01/25/copilot-research.aspx
@veronica @ajsadauskas @technology The hype around AI in software engineering seems to be that it is ‘proven’ that devs produce code quicker. it is going to be interesting to see if the corporate world values code quality over development velocity. There seems to be a pervasive belief that “move fast and break things” is how the big guys do software engineering. A few points to note:
- this idiom only applies when you fail fast, realize it, and address the problem that has been introduced.
- Break things does not mean enshittify ie create tech debt by virtue of poor code
- It really only applies if you have enough development resources to do the rework. That is to say, can afford to get it wrong often.
#AI #copilot
@ajsadauskas @technology A system that can’t balance brackets and is awful at math is gonna do great 👍
@ajsadauskas @technology …and they’ll find themselves lapped by those who didn’t drink the LLM Kool-Aid.
@ajsadauskas @technology good prediction. This is basically what they always do with every overly-hyped technology.
@ajsadauskas @technology And no one will care.
That happens right now and usually it was the CTO that wrote that shit when the company was a startup.