On Saturday, November 08, 2025, at 11:10 AM, James Bowery wrote:
> Pareto frontiers don't provide a principled loss function.  You have to 
> somehow bite the bullet and bring the complexity measures into the same units 
> and add them.

That structure wasn’t the structure I was intending but it’s interesting that 
ChatGPT was able to spit it out. I'm not sure if I agree with “AGI is the fixed 
point of self-evaluated self-rewrite.“ or “AGI is what remains when improvement 
can no longer beat itself.“. I think of AGI as being a dynamic continuously 
evolving structure integrating many Pareto frontiers. Also ChatGPT’s generated 
source code is good for examples but… I’ve tried taking it’s sample code and 
Vibing with it but it’s still tedious and too much try and fail manual 
intervention is required.

What would be nice is a cad-like interface to build the category theory and 
math and then some functions for source code generation. I’m sure that's coming 
or already exists somewhere.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T19f9264b814f3dfa-M08a6225a940af6f84443d9be
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to