[ 🏛️ new / 🏖️ lounge / 🧻 scrap ] [ 🔎 ] [ 🏠 Home ]

/lounge/ - Lounge

Ultimate Manchildren's Playpen
Name
Option
Subject
Comment
Flag
File
YouTube
Password (For file deletion.)


File: 1753407107538.jpg 29.66 KB, 522x600, free laundry.jpg

 No.103732

What are some patterns you've seen with obviously generated text that immediately make you close the article?

>"No(t)[x], no(t) [y], no(t)[z], just"


>"This isn't a/just/just a/some [x]-it's a hurrpin' derpin' doo!"


>"[usually closing alliteration where it uses three words that start with the same letter]"


I hate this shit so much, man. AI has its place (primarily in science and medicine) but people are literally replacing it with everything they do and it's just gay.

 No.103734

The extra long dash thing

 No.103735

>>103734
Oh yeah, what's up with that? It's never a normal "-" it's a character I don't even know how to recreate off of the top of my head, like something you have to pick out of the character map manually.

 No.103744

AI definitely does not have a place in medicine, because it ruins critical thinking which is really important for medicine. There are too many midwits in medicine as it is and AI makes them even worse at their job. It can be used, but only if you're not a midwit to start with and use it quite literally like a search engine.

Excessive use of emojis is another one, OP.

 No.103745

>>103744
"AI" is going to make everyone retarded in the same way nobody can spell without an automatic spellchecker anymore. Convenience kills and the only thing it's going to contribute to is brain atrophy.

 No.103746

>>103744
Most doctors are already just middlemen (aka useless parasites) between patients and google/chatGPT. The 1% of doctors that actually use their brains are the exception and not the rule. Hopefully AI fully replaces all of them

 No.103748

>>103744
I'll clarify, I only meant circumstantially where it's proven itself like with visual imaging (supposedly AI was able to catch the formation of cancer much better than the tool-assisted naked eye-but maybe that was a fluke too). Using it as a search engine has proven to be a dubious decision even in a civilian context.

 No.103753

>>103746
>Most doctors are already just middlemen (aka useless parasites) between patients and google/chatGPT. The 1% of doctors that actually use their brains are the exception and not the rule. Hopefully AI fully replaces all of them
Where LLMs are better than humans and search engines:
• memorization and breadth of knowledge
• speed at providing results
• emulating speech styles, e.g. formal corporate/resume writing

Where LLMs suck
• no deductive reasoning, hence frequent incorrect conclusions and "hallucinations"
• no numerical precision and formal computations
• non-local models are censored and politically correct
• non-local models limit you on how much of your input it can memorize in a conversation and the length of conversations ("token" limits)
• non-local models do not memorize your previous conversations beyond ChatGPT's limited "memory" feature
• it almost always tells you what you want to hear based on how you phrase the question (barring hitting any censors)
• its training/knowledge is only from written text (again much of which is politically correct, and just plain bullshit), not observable real life experience

You need more than just an LLM for "AI" to replace human doctors.

>>103745
>nobody can spell without an automatic spellchecker anymore
English is a bullshit language that we're stuck with thanks to the British Empire. A superior language would be phonetic. Nobody ever needs a dictionary or spellchecker when they write in German, Spanish, Portuguese, Italian, Latin, Greek and so on.



[Return] [Catalog] [Top][Post a Reply]
Delete Post [ ]