• Home
  • Now
  • RSS

HPSHELTON

Programming, Privacy, Politics, Photography

Sep 21, 2025

A non-anthropomorphized view of LLMs →

Instead of saying "we cannot ensure that no harmful sequences will be generated by our function, partially because we don't know how to specify and enumerate harmful sequences", we talk about "behaviors", "ethical constraints", and "harmful actions in pursuit of their goals". All of these are anthropocentric concepts that - in my mind - do not apply to functions or other mathematical objects. And using them muddles the discussion, and our thinking about what we're doing when we create, analyze, deploy and monitor LLMs.

Older →

← Newer

 

Links

  • RSS
  • GitHub
  • Liked Posts
  • LinkedIn

H. Parker Shelton

I'm just an ordinary thirty-something who's had some extraordinary opportunities. I graduated from Johns Hopkins University, work for Microsoft in Silicon Valley, code websites and applications, take the occasional photograph, and keep a constant eye on current events, politics, and technology. This blog is the best of what catches that eye.

 
  • © 2010 – Present, H. Parker Shelton (Except Where Noted)
  • Hosted by GitHub Pages
  • Design by Ian P. Hines