Brendan Greene von PlayerUnknown sagt, dass KI-Inhalte das Internet ruinieren, weil es „eine Schleife ist, LLMs scannen diesen Müll, und dann wird das zur Wahrheit … es ist wie ein Wettlauf in die Mitte der Scheiße“: „Wie kann man Dingen vertrauen, die ganz unten sagen, dass man alle Antworten, die ich gebe, einer Faktenprüfung unterziehen muss?“
https://www.pcgamer.com/software/ai/brendan-greene-says-ai-content-is-ruining-the-internet-because-its-a-loop-llms-are-scanning-this-junk-and-then-that-becomes-truth-its-like-a-race-to-the-middle-of-sh-t/
23 Kommentare
You don’t trust it. You don’t use it. That’s the only way to get these companies to understand how garbage their product is. But people want easy answers and AI is seemingly giving that to them.
There are pricing models in finance that attempt to predict the value of an asset based on certain prior information. CAPM/WACC for equities, Black-Scholes for options, as an example. They are so prevalent and “good enough” that most people accept their outputs as reality, therefore turning them into a generally accepted reality.
I believe this is what LLMs are going to do for everything. They will be good enough, accepted enough, that they will end up dictating what is real.
Ah yes, I fondly remember the pre-AI internet where everything was true and no facts needed checked.
Did he trust everything he was told before AI?
Getting real tired of endless clickbait stories where they just pick some random guy and get them to give their uninformed opinions about this stuff. Why should I care about Brendan Greene’s opinion on this? What authority does he have?
AI, the modern day pop-up, spam email, and car warranty phone calls. What a waste of good technology.
Oh its ruining so much more than just the internet. How its being used in law, medical, and now with tax season used in finances, ai is ruining *lives*. Its the greatest con we’ve been sold and nobody gets to answer for its problems because the users point to the ai and the tech bros hand wave it away through disclaimers and that terms of use you agreed to.
AI might not be as accurate as a specialist in a field, but it’s still more accurate than an average person’s conversation. Also, it never has just one response. If I ask the same question every day, the answer varies a little bit every time. I’m not sure this is actually much of an issue.
This is the idea. Fascists need it to be a contest of wills and propaganda because facts undermine fascism.
This is the Republican dream. Never forget this. One day the Democrats will pass regulations for this stuff but it’ll be too late. Republicans hate knowledge, they hate people knowing things. They want to be able to have their God-Emperor go out and shout lies from the rooftops and not give you any way to combat his misinformation. Remember who did this.
Why are LLM allowed to be trained on anything that isn’t factual? It seems rather irresponsible to let it just scan the internet willy nilly. I mean we have the younger generations doing that on their own and they wind up as misogynistic nazis over on twitter.
Wait until you learn about LLM model collapse.
Aka ‘Model Collapse’ for anyone interested (didn’t see it mentioned in the article). ✌🏽
Tailored training archives are the way to go. Software maintenance is proving that. But creating them is boring work and the superstars don’t want to do it.
It’s really only continuing the trends that were already there. People posting fake images? Pre-dates LLMs or DMs. Bots posting on social media? Pre-dates LLMs. Spreading fake news or propaganda online? Pre-dates LLMs. The internet was already ruined, LLMs just made it easier to keep it that way.
Kurtzghezacht (sp?) did a video about this AI loop a couple months ago
Just wait til you see facebook posters sharing screenshots of ai search and treating it as gospel. Granted that crowd was already cooked, but still.
I feel like this is only problem for people that were already believing everything they read on the internet instead of fact-checking. It didn’t create a problem – it exacerbated one that already existed.
You really think people would do that? Just go on the internet and tell lies?
There is a great video where a guy asks LLM-bots via voice to tell him all the numbers from 1 to 100 that are spelt with the letter „A“.
Invariably you get one clanker insisting that eight fits the bill before giving other numbers like four, forty two, thirty six, etc.
It’s embarrassing that we as a society allowed this „AI“ nonsense go on this long.
and where are you supposed to fact check things if everything is AI created?
And that’s coming from a guy who just steals ideas. Good job calling that out.
AI is becoming the Ouroboros.