From: n0t@valid.email   
      
   On 10/28/25 2:56 AM, thule wrote:   
   > i'm not one of those ai-hating redditors; i host a few small models on   
   > my home network and use them regularly.  i'm unqualified but i have read   
   > a few computational neuroscience textbooks and i'm convinced llms cannot   
   > become agi.  a fundamental limitation is their lack of world models, and   
   > some models now work around this with chain of thought, but it's not a   
   > true solution.  while general intelligence is defined as whatever humans   
   > can do, i doubt there's such thing as truly "general" - i.e. opposite of   
   > narrow - intelligence; indeed, the brain is a collection of specialised   
   > regions bootstrapped together, with the cortex and its canonical circuit   
   > being the only sort of general-purpose structure i'm aware of.   
      
   My guess is that OpenAI and others will be trying to create something   
   that works in a very similar way to the human brain when replicating AGI   
   behavior.   
      
   A single monolithic LLM is unlikely to solve the problem of AGI, but a   
   network of LLMs, each attached to its own set of long-term memories and   
   working together—might get closer, especially with orders of magnitude   
   more compute. It's the arbitrator "function" that chooses the path (akin   
   to subconscious and conscious decision-making) that complicates things.   
      
   Not saying it's easy, or even that close, but I really don't think it's   
   as unattainable as it might seem. Think Kryten or Nexus 6; start with   
   rules based directives and evolve to a more general understanding of   
   reality through lived experience.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)