Thread

1. randomImmigrant · 1789513790 · focus · HN ↗

   I think bearish on LLMs for automation, and bullish for LLM+human experts in specific fields, is about the right expectation for current architectures.

   Apart from issues with task generalization, or perhaps related to it, is the fact that LLMs have real trouble with timekeeping, and cannot estimate the real world time it will take them to do things very well. This plus the memory issues make dreams of long horizon agents, that could plausibly handle changing specifications, quite implausible with current architectures.

   In narrow domains with more deterministic outputs though, this is less of an issue, and we see multiple agents succeed much better.

   The fusion of that capacity, with humans in the loop able to better direct such agents and act as their temporal tethers, is where I think the real action will be for a while at least.

   1. handfuloflight · 1789517632 · focus · HN ↗

      > This plus the memory issues make dreams of long horizon agents, that could plausibly handle changing specifications, quite implausible with current architectures.

      Any reason why that can't be solved through context management and keep-forward scaffolding?

      1. arm32 · 1789525818 · focus · HN ↗

         Write the same sentence you just wrote back to me, but in only four words and let’s see if it has the same meaning.

         1. lantry · 1789526229 · focus · HN ↗

            "Any reason why that can't be solved through context management and keep-forward scaffolding?"

            becomes

            "load bearing context seam"

            /s

            1. bitwize · 1789554713 · focus · HN ↗

               You're gonna have to learn to talk that LLM speak!

               Dabadooba, ba dabadooba!

               <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=egpWCC2svVo" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=egpWCC2svVo

Unofficial Hacker News client; not affiliated with Y Combinator.