For several years, Schmidhuber's brother
(a theoretical physicist @ Munich, Caltech, Princeton, Bern, CERN)
lived next door to Einstein's former apartment in Bern, Switzerland.
Feedback Neural Networks,
Optimal Problem Solvers,
Reinforcement learning (RL),
Meta-Learning or learning to learn,
Artificial Curiosity & Creativity & Intrinsic Motivation & Developmental Robotics,
Theory of Beauty,
Generalized Algorithmic Information
First Schmidhuber will build a scientist better than
himself (his colleagues claim that should
be easy) who will then do the remaining work.
He cannot see any more efficient
way of using and multiplying the little creativity he has got.
This already motivated
his diploma thesis on metalearning and recursive self-improvement (1987), where
(GP) to itself, to recursively evolve
better GP methods.
This also motivated his
work on the Gödel machine (since 2003),
which formalizes I. J. Good's informal remarks (1965) on
an "intelligence explosion" through self-improving "super-intelligences".
on Schmidhuber's work (The ideal scientist, 6/2/04, 5 pages, in German).
See also the interviews in HPlus Magazine (1/5/2010): Build An Optimal Scientist, Then Retire,
and in The Next Big Future (1/18/2011)
Ich habe wenig Geduld mit Wissenschaftlern,
die sich ein Brett suchen und an der dünnsten Stelle
zu bohren anfangen.
Check out the formal theory of creativity (1990-2010)
science, art, music, humor, and describes the simple algorithmic
principles of artificial scientists & artists.
Robot Population Explosion,
Resilient Robots (Science 316 p 688),