Skip to main content

Developer crying and yelling: StackOverflow and forum sucks. Its code quality is questionable and poor.

Also, the same developer: Using ChatGPT generated code happily and telling everyone how productive he is now with it.

Where do you think ChatGPT got its data to train LLM? Do you think all of this code comes from the proprietary vendors' internal system? LOL. I tell you, some devs can be so clever but dumb simultaneously. Let us call them Schrödinger devs.

in reply to nixCraft 🐧

Nazo
Nazo
• •
I imagine they also sto- -- uh, borrowed without permission or any respect for the creator's rights -- from Github and others like it too.
in reply to Nazo

nixCraft 🐧
nixCraft 🐧
• •
Correct. Companies like ChatGPT/OpenAI, Microsoft, Google, and other LLM/ AI firms often lack strong ethical standards. Their primary focus is generating profits and pleasing their shareholders rather than respecting copyright or other ethical considerations. They be like: Why should we care about the sources or give credit for LLM-generated outputs when we can use other humans work to a train LLM and it is free labor. they don't care. It makes me sick
This entry was edited (5 days ago)
in reply to nixCraft 🐧

Nazo
Nazo
• •
It seems like if companies declare that a thing is for the sake of monetary progress, then all laws and morals go out the window.
in reply to nixCraft 🐧

ferricoxide
ferricoxide
• •
ChatGPT also scours GitHub.
in reply to ferricoxide

nixCraft 🐧
nixCraft 🐧
• •
@ferricoxide yes, not just github. if you wrote wiki entry, video, audio, or blog or book, it is all consumed by ChatGPT. Everything out there is now part of their system.
@ferricoxide
in reply to nixCraft 🐧

nixCraft 🐧
nixCraft 🐧
• •

Well-established old principle: Do one thing and do it well.

Modern AI/LLM companies' principle: Do 1000 things, and none of them are right.

Welcome to the 21st century. We have robots that can provide you any information.*

*In extremely small letters: We can make mistakes, so double-check it.

in reply to nixCraft 🐧

ElectroFetish
ElectroFetish
• •
Modern AI/LLM companies' principle: Do 1000 things, and none of them are right.
Unfortunately, this principle already works in programming: lot of programs that suffer from a lack of optimization
in reply to nixCraft 🐧

Hans van Schoot
Hans van Schoot
• •
the double-check it part is getting harder and harder every day IMHO. Search results are getting poisoned with the same flawed junk output as the LLMs are generating.
We should have archive.org-ed the Google search index database itself around 2020 or so, to make it possible to search without having to weed through the junk...
(and yes, the answer for many linux related questions can usually be found in man pages, fortunately those are still written by a Hu-man)
This entry was edited (5 days ago)
in reply to nixCraft 🐧

Kelvin Shadewing
Kelvin Shadewing
• •
No, you don't understand the power of Chat Gipity. The poor code it generates gives me something to debug. I learn from its mistakes and get better as a coder. XD
⇧