Post ARRmK2nubSrmR8SBCi by maxwainwright@toot.community
 (DIR) More posts by maxwainwright@toot.community
 (DIR) Post #ARROyQ2TEpqlBJyyYa by JamesGleick@zirk.us
       2023-01-08T16:21:24Z
       
       0 likes, 0 repeats
       
       We see that #ChatGPT can be astonishingly plausible without necessarily being accurate, factual, or “truthful.” It generates misinformation as persuasively and enthusiastically as correct information, making it difficult to distinguish truth from lies.Is this a choice by #OpenAI, or is it inevitable with the large-language-model approach?It’s all too appropriate for our time, as mendacity and bullshit-artistry poison our political discourse—demagogues thriving in a fact-free realm.
       
 (DIR) Post #ARROyQdL1lLB1fEQ1Q by clive@saturation.social
       2023-01-08T18:05:42Z
       
       0 likes, 0 repeats
       
       @JamesGleick As for choice, it's hard to say; though it's quite possible that folks in the tech sector are sufficiently surrounded by -- and acclimatized to, if not *comfortable* with -- people who confidently run their mouths about subjects they have no expertise into (a great example: the *ludicriously* wrong "bruh my off-the-cuff analysis is that COVID won't be a big deal" tech-dude post that got passed around early in the pandemic) ...... that GPT-3 truly seemed "human"
       
 (DIR) Post #ARRP94cnCD43x63Ino by clive@saturation.social
       2023-01-08T18:07:40Z
       
       0 likes, 0 repeats
       
       @JamesGleick And there's Harry Frankfurt's superb "On Bullshit", which fits ChatGPT perfectly:https://clivethompson.medium.com/on-bullshit-and-ai-generated-prose-611a0f899c5
       
 (DIR) Post #ARRmK2nubSrmR8SBCi by maxwainwright@toot.community
       2023-01-08T22:23:33Z
       
       0 likes, 0 repeats
       
       @clive @JamesGleick I have a question. You write ”I’m only going to truly rely on material I read directly myself.”. How? Can you tell your brain to “disregard the following summary”, but still read and use it? I don’t think so. Bullshit and disinformation are tricky that way, they exploits cognitive biases. And those we can’t just correct for. Better to ignore the bullshit (e.g. not expose oneself knowingly to the output of bullshit generators).
       
 (DIR) Post #ARez0jp5driBQaedai by jamescham@mastodon.social
       2023-01-15T07:20:51Z
       
       0 likes, 0 repeats
       
       @clive @JamesGleick I think there’s still a lot of work to figure out what these tools are good for. The metaphors I’ve found helpful is to think about them as being good at being 1) translation from one language to another 2) glue (social and technical) between systems and 3) paths from one conceptual space to another. But they aren’t really good for reliable information retrieval!
       
 (DIR) Post #ARfstTmiT6X55lexY8 by clive@saturation.social
       2023-01-15T17:46:19Z
       
       0 likes, 0 repeats
       
       @jamescham @JamesGleick yes, exactlyPersonally, I was really looking forward to using them for summarization -- i.e. "give me a two-paragraph precis for each o these 10,000-word white papers so I can figure out which ones I should read in full" ...... but I do worry that the accuracy of summaries won't be reliable