Do you want to get kidnapped and handcuffed to a radiator for 24 hours? Because that's how you get kidnapped and handcuffed to a radiator for 24 hours.
what's the market on viking_hippie getting kidnapped and getting handcuffed to a radiator for 24 hours because let me know when that one starts to go up
That was always my assumption of the end game. You have the system prompts, an advertising bias prompt layer over top, then the user prompts.
"Naturally worded" advertising that doesn't immediately appear to be advertising and searching using natural language always seemed to be the biggest use cases for LLMs to me, considering they can't be relied on to output accurate info.
Exactly. Like big pharma paying ChatGPT to convince you that your symptoms are an illness they have pills for. It'd turn the LLMs from librarians into salesmen playing librarians lol.
Yes, I’ve been saying this for over a year now about AI-driven search. And not just specifically search engines. Like “where’s the nearest coffee shop?” To your phones assistant or whatever
I currently tend toward perplexity because traditional search engines are so terrible now thanks to ads and seo. We’re speedrunning llms being just as useless at finding things
Have they rolled out the ads yet ? I feel like they keep teasing it but I might have missed the roll out and the backlash (there is no way there would not be SOME backlash)
LLM output is one of the few places I’d be OK with seeing ads. I’d almost go so far as advocating for it to be mandated, because it would help make slop easier to identify and filter out.
The real issue is that since any fingerprint that can be mandated for AI content must be algorithmically implemented, then that fingerprint can be algorithmically removed.
For example, let's say companies voluntarily choose or are forced to integrate text fingerprinting into LLM output. Automated AI writing detection tools already exist, but they're not reliable. But in principle we could make the output of LLMs easy to identify. Maybe we force them to adopt subtle but highly unique patterns of word choice, punctuation, sentence structure, etc. Then if any student attempted to upload an LLM-generated essay to their course website, the system could with high accuracy flag it as AI generated.
But...if those patterns are so clear and unambiguous, it also means they can be easily detected by third party tools. If one person can code ChatGPT to add special fingerprinting to the text ChatGPT creates, another person can create a program that you can paste ChatGPT text into that will remove that fingerprinting.
29 Comments
ivan@piefed.social · 87 pts · 98d
As someone doing teaching, I already had quite enough of "If you have any further questions - just let me now! 😊" in answer boxes.
Kanda@reddthat.com · 46 pts · 98d
Easy F for AI and move on to the next one
Flower@sh.itjust.works · 11 pts · 98d
Remember the time every answer started with "As a large language model..."
definitemaybe@lemmy.ca · 4 pts · 98d
When it was new, I got a lot of "Certainly! ..."
Like, come on, mate. You didn't even read the first sentence before submitting.
howrar@lemmy.ca · 1 pts · 98d
Sounds like a challenge
massive_bereavement@fedia.io · 75 pts · 98d
Fuck polymarket though, the metastasis of late stage capitalism.
Viking_Hippie@lemmy.dbzer0.com · 32 pts · 98d
I'm gonna bet everything I own that Viking_Hippie plays a video game within the next 24 hours. Any takers?
gnutrino@programming.dev · 36 pts · 98d
Do you want to get kidnapped and handcuffed to a radiator for 24 hours? Because that's how you get kidnapped and handcuffed to a radiator for 24 hours.
HeyThisIsntTheYMCA@lemmy.world · 15 pts · 98d
what's the market on viking_hippie getting kidnapped and getting handcuffed to a radiator for 24 hours because let me know when that one starts to go up
youcantreadthis@quokk.au · 3 pts · 98d
My brain runs DOOM, or so I'm told.
IAmNorRealTakeYourMeds@lemmy.world · 19 pts · 98d
It's a form of psychosis.
they think it predicts the future, and so must be completely unregulated.
They created an Oracle made of gold.
rapchee@lemmy.world · 3 pts · 98d
i haven't seen those guys much but based on the little i've seen, i think they believe whatever makes the most amount of money
SubArcticTundra@lemmy.ml · 58 pts · 98d
Just wait till they let advertisers buy hidden bias in its responses
wizardbeard@lemmy.dbzer0.com · 44 pts · 98d
That was always my assumption of the end game. You have the system prompts, an advertising bias prompt layer over top, then the user prompts.
"Naturally worded" advertising that doesn't immediately appear to be advertising and searching using natural language always seemed to be the biggest use cases for LLMs to me, considering they can't be relied on to output accurate info.
SubArcticTundra@lemmy.ml · 17 pts · 98d
Exactly. Like big pharma paying ChatGPT to convince you that your symptoms are an illness they have pills for. It'd turn the LLMs from librarians into salesmen playing librarians lol.
SaraTonin@lemmy.world · 9 pts · 98d
Yes, I’ve been saying this for over a year now about AI-driven search. And not just specifically search engines. Like “where’s the nearest coffee shop?” To your phones assistant or whatever
I currently tend toward perplexity because traditional search engines are so terrible now thanks to ads and seo. We’re speedrunning llms being just as useless at finding things
WhatAmLemmy@lemmy.world · 5 pts · 98d
That's... exactly how they implement advertising in an LLM?
SubArcticTundra@lemmy.ml · 1 pts · 98d
sobchak@programming.dev · 2 pts · 98d
It's already pretty biased. For web projects it almost always recommends Supabase and Vercel.
SubArcticTundra@lemmy.ml · 1 pts · 97d
I mean when they do introduce intentional bias, we'll have no way of proving it for certain
njm1314@lemmy.world · 17 pts · 98d
Why is Polymaket breaking news?
sobchak@programming.dev · 3 pts · 98d
market manipulation
rem26_art@fedia.io · 13 pts · 98d
is polymarket reporting news here or are they reporting the results of a
betevent contract?Admax@lemmy.world · 12 pts · 98d
Have they rolled out the ads yet ? I feel like they keep teasing it but I might have missed the roll out and the backlash (there is no way there would not be SOME backlash)
nonentity@sh.itjust.works · 11 pts · 98d
LLM output is one of the few places I’d be OK with seeing ads. I’d almost go so far as advocating for it to be mandated, because it would help make slop easier to identify and filter out.
isleepinahammock@lemmy.blahaj.zone · 3 pts · 98d
The real issue is that since any fingerprint that can be mandated for AI content must be algorithmically implemented, then that fingerprint can be algorithmically removed.
For example, let's say companies voluntarily choose or are forced to integrate text fingerprinting into LLM output. Automated AI writing detection tools already exist, but they're not reliable. But in principle we could make the output of LLMs easy to identify. Maybe we force them to adopt subtle but highly unique patterns of word choice, punctuation, sentence structure, etc. Then if any student attempted to upload an LLM-generated essay to their course website, the system could with high accuracy flag it as AI generated.
But...if those patterns are so clear and unambiguous, it also means they can be easily detected by third party tools. If one person can code ChatGPT to add special fingerprinting to the text ChatGPT creates, another person can create a program that you can paste ChatGPT text into that will remove that fingerprinting.
pewgar_seemsimandroid@lemmy.blahaj.zone · 7 pts · 98d
they stealing from youtubers again.
IAmNorRealTakeYourMeds@lemmy.world · 5 pts · 98d
Tbf, a student including relevant info they leaned from world of tanks wouldn't be a bad thing.
daniskarma@lemmy.dbzer0.com · 16 pts · 98d
Then there's the student who plays wanthunder and leaks a classified F-47 detailed blueprint onto the paper.
IAmNorRealTakeYourMeds@lemmy.world · 10 pts · 98d
that deserves extra credit