Mercurial > cortex

comparison thesis/abstract.org @ 436:853377051f1e

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
abstract v. 2
author Robert McIntyre <rlm@mit.edu>
date Sun, 23 Mar 2014 19:09:14 -0400
parents ae3bfc82ac7c
children c1e6b7221b2f
comparison
equal deleted inserted replaced
435:ae3bfc82ac7c 436:853377051f1e
1 Here I explore the design and capabilities of my system (called 1 Here I demonstrate the power of using embodied artificial intelligence
2 =CORTEX=) which enables experiments in /embodied artificial 2 to attack the /action recognition/ problem, which is the challenge of
3 intelligence/ -- that is, AI which uses a physical simulation of 3 recognizing actions performed by a creature given limited data about
4 reality accompanied by a simulated body to solve problems. 4 the creature's actions, such as a video recording. I solve this problem
5 in the case of a worm-like creature performing actions such as curling
6 and wiggling.
5 7
6 In the first half of the thesis I describe the construction of 8 To attack the action recognition problem, I developed a computational
7 =CORTEX= and the rationale behind my architecture choices. =CORTEX= is 9 model of empathy which allows me to use simple, embodied
8 a complete platform for embodied AI research. It provides multiple 10 representations of actions (which require rich sensory data), even
9 senses for simulated creatures, including vision, touch, 11 when that sensory data is not actually available. The missing sense
10 proprioception, muscle tension, and hearing. Each of these senses 12 data is ``imagined'' by the system by combining previous experiences
11 provides a wealth of parameters that are biologically 13 gained from unsupervised free play.
12 inspired. =CORTEX= is able to simulate any number of creatures and
13 senses, and provides facilities for easily modeling and creating new
14 creatures. As a research platform it is more complete than any other
15 system currently available.
16 14
17 In the second half of the thesis I develop a computational model of 15 In order to build this empathic, action-recognizing system, I created
18 empathy, using =CORTEX= as a base. Empathy in this context is the 16 a program called =CORTEX=, which is a complete platform for embodied
19 ability to observe another creature and infer what sorts of sensations 17 AI research. It provides multiple senses for simulated creatures,
20 that creature is feeling. My empathy algorithm involves multiple 18 including vision, touch, proprioception, muscle tension, and
21 phases. First is free-play, where the creature moves around and gains 19 hearing. Each of these senses provides a wealth of parameters that are
22 sensory experience. From this experience I construct a representation 20 biologically inspired. =CORTEX= is able to simulate any number of
23 of the creature's sensory state space, which I call \phi-space. Using 21 creatures and senses, and provides facilities for easily modeling and
24 \phi-space, I construct an efficient function for enriching the 22 creating new creatures. As a research platform it is more complete
25 limited data that comes from observing another creature with a full 23 than any other system currently available.
26 compliment of imagined sensory data based on previous experience. I
27 can then use the imagined sensory data to recognize what the observed
28 creature is doing and feeling, using straightforward embodied action
29 predicates. This is all demonstrated with using a simple worm-like
30 creature, and recognizing worm-actions in video.
31
32
33 empathy is important
34 cortex tests that idea