Yui vs Alice Light novels Spoiler :P
anyway, joke aside,
Yui is a self aware Top-down AI.
The Self awareness make cloning her a problem as it can cause problem of not being unique and end up breaking down or wanting to destroy the other one.
Also, The reason Yui is so human like, it's because she simply have a huuuuuuuuuuuuge amount of data sample to make her decisions from and act like humans.
That huge amount of data is actually her monitoring and analyzing the memory and emotions of 10000 players in Aincrad for 2 whole years.
You could say Aincrad and the Cardinal system, had a secret mission inside them that created human-like AI by all the data they analyzed from the players in it all living in aincrad like their 2nd life and reality.
this was hinted and shown many many time in the novels, sadly the anime only would not catch it easily.
But it will be more and more apparent later on as well, specially if we get a Progressive anime or a continuation after Alicization into Unital Ring Arc.
Anyway, back to the main topic
if Rath really want, they can copy yui and then remove parts of her personality to create some clones that are not really yui but still can grow to be similar human like AI.
heck, even ALO's Cardinal is able to create some very human like AI. ( Only ALO's Cardinal which is full version of Cardinal, The Seed version of Cardinal is not full version )
But the problem with having a huge number of solider... all using the same code is HUUUUUUUUUUUUGE.
for one, they will all end up having similar respond to similar situations, that mean their decision making pattern can be studied and guessed by the enemy and easily countered.
the 2nd reason would be that they can't react to unpredictable stuff if it's completely out of their database of decision making algorithms and in the middle of the war with signal jamming and other stuff, they can lose connection to their base and get isolated.
the last and most important one is....
Since they all would have the same algorithm, after a while the enemy can finally hack them, and hacking one, mean easily hacking all the others too over time, that mean the whole AI Solider system would be compromised.
but non of this is a problem for Bottom-up AI, as they don't work over some algorithm, they can't be hacked and can only be deceived, like a normal human, but as long as they are trained well and matured, they clearly won't be deceived and can fight like any other normal human solider.
And this is Rath's goal !
Plus, the fact that they can create veteran solider and generals, all of which with their own battle experience and their own unique decision making in a very short time and with a very low cost is a huuuuge pay off.
if you didn't get what I mean, I explain in a short example :
1. Create a new Modern UnderWorld with guns and countries.
3. Increase the UnderWorld Speed to 5000x and w8 ( 1 day = ~13 years in UW )
2. Let them grow and start a war between countries. similar to Final Load Testing they have in UW.
3. they will fight and the survivor of the army all will be veteran war general and soldiers with their own war experience.
4. Take them out and use them as Solider in real world.
See... Easy creation of war veterans without the use of any time or even any cost of training !!!! |