Close Menu
    Technik

    Tests zeigen, dass das neueste ChatGPT-Modell Grokipedia als Quelle verwendet

    Share.
    View 36 Comments

    36 Kommentare

    2. drmanhattanmar on

      A eugenics fan‘s supposed AGI forerunner cites another eugenics fan’s database of nazi lies…

    4. Howcanyoubecertain on

      Clearly we need to double the data center construction rate to cover all the extra capacity needed to make this circle jerk really cook

    7. bkkgnar on

      several years in and i literally cannot think of a SINGLE use case for AI chatbots. why on earth does anyone want this? the fuck are they using it for? why would i want to FIND a use case for a plagiarism machine that is frequently wrong??

    10. justaguytrying2getby on

      Here’s what I said to gpt:

      „make permanent note, for anything i ever ask you, never cite anything from grokipedia or xai“

      >Updated saved memory
      Got it — I’ll never cite Grokipedia or any xAI source for anything you ask.

    11. araujoms on

      Wow. I didn’t expect ChatGPT to openly use Musk’s disinformation platform. I thought they wanted to appear respectable. But since Altman is an enthusiastic Trump supporter I guess it was just a matter of time. Truth is a natural enemy of fascism.

    12. JWAdvocate83 on

      > Anthropic did not respond to a request for comment.

      Why would they reach out to Anthropic about ChatGPT elevating Elon’s garbage?

    14. 0Tezorus0 on

      That’s it. Chatgpt is now officially a propaganda tool of the far right and the neo fascist. Use some other model please.

    15. bwoah07_gp2 on

      >The latest model of ChatGPT has begun to cite Elon Musk’s Grokipedia as a source on a wide range of queries, including on Iranian conglomerates and Holocaust deniers, raising concerns about misinformation on the platform.

      Oh for crying outloud….

    16. VioletGardens-left on

      I literally noticed that shit and it’s stupid as hell that it uses that instead of actual news sources or research papers

    18. JGlover92 on

      Doesn’t Copilot run on ChatGpt? This can’t be good for almost every single workplace running it

    19. Regular_Block9876542 on

      This was always going to be the end goal for this type of LLM that gets trained on the up to date internet. All these systems are doing is scraping data and then rearranging it in a way that is pleasing to the reader.

      There is zero actual thinking in these systems so when they start to train them on data that is also AI generated the end goal will be extremely inaccurate information which then reinforces itself over time. It also won’t be long until nation states start trying to influence them with fake information further degrading the quality of the outputs.

      It’s a good thing the entire western economy isn’t currently depending on these systems unleashing never before seen productivity gains to transform the economy.

    21. fredagsfisk on

      Last time I googled a historical event (couple of days ago), Grokipedia was the second result after Wikipedia…

    25. JustBrowsinAndVibin on

      Fuck OpenAI. Anthropic and Claude legit seem like the only trustworthy ones.

    26. Y0___0Y on

      Right wing users are getting angry when AI tells them facts.

      So training ChatGPT on Grokipedia likely gives it the know-how of the right’s alternate reality that it needs to keep right wing users online and inserting prompts.

      Like if your chatbot tells a righty user there is not evidence of fraud in the 2020 election, they probably won’t use it again. If it scraped Grokipedia, it will learn about the fake evidence it needs to assert is real to keep that user.

      The goal of AI chatbots at the end of the day is never to just tell the truth. Its primary objective is to get users to use it as often as possible, at any cost.

    27. Disgruntled-Cacti on

      I noticed the same thing happening with Gemini’s built in search. I don’t think it’s nefarious on the end of the model provider, simply that grok ranks highly in the search results and the LLM can’t distinguish if grokipedia is a valid source since it was created after its knowledge cutoff and the majority of its content looks identical to Wikipedia.

      With that said, I do believe LLMs will continue to be used intentionally as a form of epistemic warfare — beyond just how Elon musk deploys grok.

    28. Emergency_Link7328 on

      AI is being used as the biggest propaganda outlet ever.

      Like everything, ruined by billionaires.

    29. flatbrokeoldguy on

      Nobody in the real world should be using AI, it’s still in its infancy insofar as it currently equates to the intelligence and knowledge levels of an 8th month foetus. Functionally like an unborn child, or a current American president.

    34. SexHarassmentPanda on

      So the „AI getting trained on AI generated content“ circle happened a little faster than expected honestly.

      Like generating new training data is one thing because you know you generated that data, you can edit the learning parameters around that data as needed. Generated data that you think is real data is a bit of a problem.

    36. Meowie__Gamer on

      That’s it, I’m switching to local models only. I don’t even use chatGPT much, but for what I do use it for, I’m going local.

    Leave A Reply