“Hemispheres stay sane only with structured outputs and a non-llm reducer”. — cursor agent. I’ve been deep in the research around “AI” or LLM’s and trying my best to figure out where it applies in critical workflows and processes. The fact that the larger public still sees it as intelligence at all is a testament to the marketing. Especially because nothing novel that couldn’t have been done through sheer repetition has been invented. Still the split brain problem intrigues me, because the research around it seemingly only considers it a binary problem. In reality a split brain is objectively a split brain but for an agent with multiple context sources that may contradict themselves, it’s more than just 2 hemispheres. Imagine one side of your brain says the sky is blue and the other says it’s orange, now add 6-7 more different context sources and you need a categorical system to choose a “winner”. Except those 6-7 sources could be just as right as the next source depending on almost an infinite number of facets. Without intelligence itself sanity is not achievable alone. Excited to see the math behind whatever solves this problem #generativeAI
Controlling the chaos of systems and subsequently people in organizations, workflows and processes. Trained by RTFM and spectacle failure at scale.
Currently: kids, wife, engineering, homesteading
Pronoun: human
| musings | https://cwarner.kernelcode.com |
| radio | KD2RYT |
| Gravatar | https://gravatar.com/christophwarner |
nixCraft 🐧
Erik Jonker
wilander
Colin Ventura
