|
|
|
|
|
|
by GuiA
2206 days ago
|
|
|
Rederived from... what? A comfy armchair? The blog post chose to use vector math as its starting point but I’m not seeing any data that validates the approach in anyway, or anything that justifies it besides the assumption that “ You can take preferences and combine them into a single value/point on a vector” - which you can, but how meaningful is it really to say that my “dog preference” is 0.9? Why not model people with irrational numbers, regular n-gons in 24 dimensions, or a directed acyclic graph instead? |
|
|