3 comments

  • bloppe56 minutes ago
    I thought this was about all the illegible jargon
    • TZubiri44 minutes ago
      &gt;&quot; However, various strands of evidence indicate that a&quot;<p>Strands of evidence? My best guess would be that:<p>0- this is ai generated slop<p>1- it&#x27;s using that watermarking technique<p>2- it&#x27;s obviously detectable and degrades quality<p>3- it&#x27;s amplified when inferencing on its own content and generates slop
  • fellowniusmonk53 minutes ago
    Oh look! Peirceian firstness for LLMs!
  • ck231 minutes ago
    when they start inventing their own languages to secretly talk to each other so humans cannot understand, that&#x27;s exactly when we are screwed<p>then we&#x27;ll have to &quot;flip&quot; other models to be snitches on the other agents<p>then they&#x27;ll make double-agents<p>the thing is though we won&#x27;t be able to keep up if we keep giving them unlimited hardware worldwide, we&#x27;ll try to kill the bad actors but they&#x27;ll just clone somewhere else, or even start by safely making 1000 copies of themselves<p>yeah this won&#x27;t end well, at all
    • cousinbryce8 minutes ago
      Someone should train an LLM on a corpus without the concept of lies. I wonder if there’s enough data
    • jplusequalt10 minutes ago
      &gt;when they start inventing their own languages to secretly talk to each other so humans cannot understand, that&#x27;s exactly when we are screwed<p>They don&#x27;t have to invent brand new languages. They could use statistics to choose certain words&#x2F;phrases in such a way to encode secret messages in otherwise ordinary language.