rlm@425
|
1 #+title: =CORTEX=
|
rlm@425
|
2 #+author: Robert McIntyre
|
rlm@425
|
3 #+email: rlm@mit.edu
|
rlm@425
|
4 #+description: Using embodied AI to facilitate Artificial Imagination.
|
rlm@425
|
5 #+keywords: AI, clojure, embodiment
|
rlm@422
|
6
|
rlm@435
|
7 -- show hand
|
rlm@435
|
8
|
rlm@435
|
9 * Embodiment is a critical component to Intelligence
|
rlm@435
|
10
|
rlm@435
|
11 * To explore embodiment, we need a world, body, and senses
|
rlm@435
|
12
|
rlm@435
|
13 * Because of Time, simulation is perferable to reality
|
rlm@435
|
14
|
rlm@435
|
15 * Video game engines are a great starting point
|
rlm@435
|
16
|
rlm@435
|
17 * Bodies are composed of segments connected by joints
|
rlm@435
|
18
|
rlm@435
|
19 * Eyes reuse standard video game components
|
rlm@435
|
20
|
rlm@435
|
21 * Hearing is hard; =CORTEX= does it right
|
rlm@435
|
22
|
rlm@435
|
23 * Touch uses hundreds of hair-like elements
|
rlm@435
|
24
|
rlm@435
|
25 * Proprioception is the force that makes everything ``real''
|
rlm@435
|
26
|
rlm@435
|
27 * Muscles are both effectors and sensors
|
rlm@435
|
28
|
rlm@435
|
29 * =CORTEX= brings complex creatures to life!
|
rlm@435
|
30
|
rlm@435
|
31 * =CORTEX= enables many possiblities for further research
|
rlm@435
|
32
|
rlm@435
|
33 * =CORTEX= User Guide
|
rlm@435
|
34
|
rlm@435
|
35 * Empathy in a simulated worm
|
rlm@435
|
36
|
rlm@435
|
37 * Embodiment factors action recognition into managable parts
|
rlm@435
|
38
|
rlm@435
|
39 * Action recognition is easy with a full gamut of senses
|
rlm@435
|
40
|
rlm@435
|
41 * Digression: bootstrapping with multiple senses
|
rlm@435
|
42
|
rlm@435
|
43 * \Phi-space describes the worm's experiences
|
rlm@435
|
44
|
rlm@435
|
45 * Empathy is the process of tracing though \Phi-space
|
rlm@435
|
46
|
rlm@435
|
47 * Efficient action recognition via empathy
|
rlm@435
|
48
|
rlm@435
|
49 * Contributions
|
rlm@435
|
50
|
rlm@435
|
51
|
rlm@432
|
52 * Vision
|
rlm@422
|
53
|
rlm@432
|
54 System for understanding what the actors in a video are doing --
|
rlm@432
|
55 Action Recognition.
|
rlm@432
|
56
|
rlm@432
|
57 Separate action recognition into three components:
|
rlm@422
|
58
|
rlm@432
|
59 - free play
|
rlm@432
|
60 - embodied action predicates
|
rlm@432
|
61 - model alignment
|
rlm@432
|
62 - sensory imagination
|
rlm@422
|
63
|
rlm@432
|
64 * Steps
|
rlm@432
|
65
|
rlm@432
|
66 - Build cortex, a simulated environment for sensate AI
|
rlm@432
|
67 - solid bodies w/ joints
|
rlm@432
|
68 - vision
|
rlm@432
|
69 - touch
|
rlm@432
|
70 - vision
|
rlm@432
|
71 - hearing
|
rlm@432
|
72 - proprioception
|
rlm@432
|
73 - muscle contraction
|
rlm@425
|
74
|
rlm@432
|
75 - Build experimental framework for worm-actions
|
rlm@432
|
76 - embodied stream predicates
|
rlm@432
|
77 - \phi-space
|
rlm@432
|
78 - \phi-scan
|
rlm@425
|
79
|
rlm@432
|
80 * News
|
rlm@432
|
81
|
rlm@432
|
82 Experimental results:
|
rlm@425
|
83
|
rlm@432
|
84 - \phi-space actually works very well for the worm!
|
rlm@432
|
85 - self organizing touch map
|
rlm@425
|
86
|
rlm@425
|
87
|
rlm@432
|
88 * Contributions
|
rlm@432
|
89 - Built =CORTEX=, a comprehensive platform for embodied AI
|
rlm@432
|
90 experiments. Has many new features lacking in other systems, such
|
rlm@432
|
91 as sound. Easy to model/create new creatures.
|
rlm@432
|
92 - created a novel concept for action recognition by using artificial
|
rlm@432
|
93 imagination.
|
rlm@426
|
94
|
rlm@426
|
95
|
rlm@432
|
96
|
rlm@432
|
97
|
rlm@432
|
98
|
rlm@432
|
99
|
rlm@432
|
100
|