Description du poste
Epistemic status: This is initial thinking, and I'm writing here to have your opinions. It's based on a recent case in which the model rejected the principle with an illogical justification. I may be wrong to assume there's a deliberate intention behind it; it could simply be a deterioration in the context. My stake in the outcome: I rely directly on controlling the generation contexts.
ContextHello everyone, I'm Abdallrahman, and this is my first post writing on LessWrong (actually the second, after rejecting the first one because of AI writing).
I started this journey about building Sijal; that institutional reasoning engine for high-stakes decisions, from a formal brainstorming after hearing Dr.Jordan Peterson call about building a system that connect others. After starting and building the first shape of Sijal, I figured out that there's more need for a decision-making process on the institutional side, reversing from 80% of the failures of institutions' expansion projects in the Middle East (even having strong capabilities!), comes from bad decisions or not testing the full strength of the opposite case or worst-case scenario.
So, I figured out the Decision making philosies, Starting with Robert Rules of order, and after studing them well, I got to belive that there's should be a rules for debating to get the best decision, I thought that this would give a high probability decsion, but it produces more than that !, I tested the system on real previous failed cases, with locking the resources (of course without any of AI thinking !) and went to the side of making it correct with these parameters - the one that available in the case time - and more to be convenced that debating systems are more logical to produce the best decisions.
On the other side, I started to use Claude Code for building the code faster; that was the cliff! Faster, Easier, More Perfect , etc., and it was driven by the current situation in Gaza and the financial problems I'm facing.
It went