22. Flight circa 1900
replicate natural phenomena
high level of public excitement
23. Flight circa 1900
replicate natural phenomena
high level of public excitement
many approaches and teams
24. Flight circa 1900
replicate natural phenomena
high level of public excitement
many approaches and teams
government and private funding
25. Flight circa 1900
replicate natural phenomena
high level of public excitement
many approaches and teams
government and private funding
lots of crackpots
26. Flight circa 1900
replicate natural phenomena
high level of public excitement
many approaches and teams
government and private funding
lots of crackpots
experts arguing it can’t be done
27. Flight circa 1900
replicate natural phenomena
high level of public excitement
many approaches and teams
government and private funding
lots of crackpots
experts arguing it can’t be done
28. 20 11
Flighti rc a 1900
circa
A Ic
replicate natural phenomena
high level of public excitement
many approaches and teams
government and private funding
lots of crackpots
experts arguing it can’t be done
29. {
Movement
Flapping Gliding
Freq. Amp. Pos.
Body
Head Wings Tail
Flight Bones
Muscles
Feathers
Contour Down Bristle
Vane Barbule Barbicels
49. Neocortex works on
several different
modalities
+
Neocortex learns
efficiently
50. Neocortex works on
several different
modalities
+
Neocortex learns
efficiently
+
Necortical
structure is
largely uniform
51. Neocortex works on
several different
modalities
A common
+
assumption-set
Neocortex learns works well
efficiently for a large set of
problems
+
Necortical
structure is
largely uniform
75. Neocortex
Source of assumptions/
constraints
Physics of World’s Data
To find correspondence
with neocortex properties
76. Neocortex
Source of assumptions/
constraints
Computational Framework
Physics of World’s Data
Understand why neocortex
To find correspondence
does what it does to design
with neocortex properties
algorithms
77.
78. Observed hierarchy
in the cortex
Hierarchical Efficiency and re-use.
structure of data Dynamic programming.
96. Engineering process
Biological circuit Explanation for
Biological biological
model for
Constraints phenomena
inference
Non-biological Predictions
learning algorithm
Hard Mathematical
Computational Empirical
Problem model for
Requirements results
inference
97. Engineering process
Biological circuit Explanation for
Biological biological
model for
Constraints phenomena
inference
Non-biological Predictions
learning algorithm
Hard Mathematical
Computational Empirical
Problem model for
Requirements results
inference
122. We are building a vision system first...
because, higher level reasoning requires
grounding in a perception-action system.
123.
124. A) After wading barefoot in the lake,
Erik used his shirt to dry his feet.
125. A) After wading barefoot in the lake,
Erik used his shirt to dry his feet.
B) After wading barefoot in the lake,
Erik used his glasses to dry his feet.
126. A) After wading barefoot in the lake,
Erik used his shirt to dry his feet.
B) After wading barefoot in the lake,
Erik used his glasses to dry his feet.
Watson?
127. A) After wading barefoot in the lake,
Erik used his shirt to dry his feet.
B) After wading barefoot in the lake,
Erik used his glasses to dry his feet.
Watson?