[HN Gopher] Large Language Models Think Too Fast to Explore Effe...
       ___________________________________________________________________
        
       Large Language Models Think Too Fast to Explore Effectively
        
       Author : bikenaga
       Score  : 57 points
       Date   : 2025-01-31 16:26 UTC (6 hours ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | Jimmc414 wrote:
       | Maps well to Kahneman's "Thinking Fast and Slow" framework
       | 
       | system 1 thinking for early layer processing of uncertainty in
       | LLMs. quick, intuitive decisions, focuses on uncertainty, happens
       | in early transformer layers.
       | 
       | system 2 thinking for later layer processing of empowerment
       | (selecting elements that maximize future possibilities).
       | strategic, deliberate evaluation, considering long-term
       | possibilities, happens in later layers.
       | 
       | system 1 = 4o/llama 3.1
       | 
       | system 1 + system 2 = o1/r1 reasoning models
       | 
       | empowerment calculation seems possibly oversimplified - assumes a
       | static value for elements over a dynamic context-dependent
       | empowerment
       | 
       | interesting that higher temperatures improved performance
       | slightly for system 1 models although they still made decisions
       | before empowerment information could influence them
       | 
       | edit: removed the word "novel". The paper shows early-layer
       | processing of uncertainty vs later-layer processing of
       | empowerment.
        
         | optimalsolver wrote:
         | > quick, intuitive decisions, focuses on what's novel/unknown
         | 
         | Wouldn't it be the exact opposite? That is, novel stimuli
         | requires more extensive processing at higher levels of
         | cognition.
        
           | hinkley wrote:
           | Requires != receives. Knee-jerk reactions to novel
           | circumstances get people into trouble all the time.
           | Particularly in impulsive people, who find their actions
           | sometimes don't align with their values, to the point they
           | can't be friends with people who are particularly judgemental
           | about actions versus words and won't accept apologies.
           | 
           | It is a lot of work to retrain your lizard brain to not react
           | in unpleasant ways when the shit hits the fan or you're
           | exhausted, and the default answer to everything is "no".
           | There are ways I show up in a crisis that are exemplary, and
           | others that I'm not proud of. Understanding why doesn't fix
           | it. It's just step 1, like in AA.
        
         | hinkley wrote:
         | Some people have been trying to claim this book is based on
         | faulty research but I definitely know people who don't believe
         | they work the way the book describes but the only sane
         | explanation for their behavior is that either they do or I know
         | an inordinate number of pathological liars.
         | 
         | And I know some neurodiverse people who absolutely believe this
         | is why they are like this. And that's always a problem with
         | studies. Some things are 10% of the time, it works every time.
         | If you test the general public for mental acuity and caffeine
         | you're going to see undiagnosed ADHD people benefitting, or
         | find they're already consuming alarming quantities of the stuff
         | (self medication).
        
         | fudged71 wrote:
         | Stanovich proposes a three tier model
         | http://keithstanovich.com/Site/Research_on_Reasoning_files/S...
        
         | svnt wrote:
         | Predicting language tokens but without logic probably does not
         | map well at all to system 1 thinking, except where
         | fast=conversation.
        
           | Jimmc414 wrote:
           | the system 1/2 analogy is obviously imperfect for token
           | prediction, but the paper does provide evidence that
           | uncertainty processing in early layers (layer 2) affects
           | decisions before empowerment processing in later layers
           | (layer 72) can influence them. quick decisions based on
           | uncertainty before deeper strategic evaluation does echo how
           | fast thinking can override slow thinking, even if the
           | mechanisms are substantially different.
        
       | kadushka wrote:
       | From the abstract:
       | 
       |  _Results show most LLMs underperform compared to humans, except
       | for the o1 model_
        
         | GaggiX wrote:
         | It's the only reasoning model they evaluated to be honest.
        
       | xerox13ster wrote:
       | Maybe this is a really really, really, really, really dumb idea:
       | 
       | What if we started Refining and reinforcing these Extremely fast
       | thinking, models with text written exclusively by stoners?
       | 
       | Stoners think slower, but they have more wide, branching
       | thoughts. I find I do some of my most seemingly creative and
       | explorative thinking when I am high.
       | 
       | If there are effects of time and season and energy in models of
       | LLMS such that you can say it's June and it performs worse or
       | it's January and it performs better. Or you can say you've got a
       | good nights rest so do better and it does, due to the fact that
       | human writing is affected by our mental state, and our mental
       | state is affected by physiological factors.
       | 
       | So then, I wonder what happens if you do HFRL and refinement on
       | text from people under the influence of marijuana.
        
         | cab11150904 wrote:
         | God I hate potheads. It's like they beg you to dislike them.
        
       ___________________________________________________________________
       (page generated 2025-01-31 23:00 UTC)