Main Point:One single “Master Algorithm” for all kinds of knowledge from chess to Go to Medicine to Stocks to …, only differ in ‘Data’ context, but the same One thinking “Mind”.
《The Master Algorithm》must have these 5 features:
- Evolving-structure-based(ala genetic),
- Learning parameters (connected) ,
- Probability-based (Baeysian),
- Symbolic (knowledge representation as symbols),
- Analogy-based (recognise similarities).
- Symbolic like “equation” used in Physics: F = ma => F – ma = 0 => general equation “U (X) = 0”
- Evolutionist: rules create new rules and discard old irrelevant rules, like the nature selection “survival the fittest “.
- Overfitting Data: Bias vs Variance. Eg. A clock aways late by 1 hour (high B, low V); if it tells almost the right time but alternate erractically between fast & slow (low B high V)
Induction is the inverse of Deduction: Deduction:
A is a, (A = Socrates, a = human) if…
View original post 200 more words