Conversation

Notices

1. Embed this notice
   Jason Gorman (jasongorman@mastodon.cloud)'s status on Friday, 29-Aug-2025 16:59:00 JST Jason Gorman

   in reply to

   LLMs are a dead end. Not reliable enough for most use cases, and never likely to be much more reliable.

   A human with expertise will always be required to instruct and monitor the work. And, as we're seeing, the productivity gains just seem to evaporate when they're actually measured because these models tend to create more downstream work than they save us.

   • Embed this notice
     Jason Gorman (jasongorman@mastodon.cloud)'s status on Friday, 29-Aug-2025 16:59:01 JST Jason Gorman

     Why do I believe LLMs are about as reliable as they're going to get? Physics, my dear chums. Physics.

     "For the case of α=0.1, reducing the error by a factor of 10 requires the allocation of ten billion more compute resources. And, if we focus on the power consumption itself, the exponent is 0.05, meaning an astronomical 10^20 more computing power for just one order better in accuracy."

     https://arxiv.org/html/2507.19703v1

     Steve's Place repeated this.