Mercurial > cortex

annotate org/intro.org @ 258:f4b67005b702

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
reorganized proprioception
author Robert McIntyre <rlm@mit.edu>
date Tue, 14 Feb 2012 00:39:16 -0700
parents 16cbce075a0b
children 7e7f8d6d9ec5
rev   line source
rlm@22 1 #+title: Simulated Senses
rlm@22 2 #+author: Robert McIntyre
rlm@22 3 #+email: rlm@mit.edu
rlm@22 4 #+description: Simulating senses for AI research using JMonkeyEngine3
rlm@34 5 #+keywords: Alan Turing, AI, sinulated senses, jMonkeyEngine3, virtual world
rlm@22 6 #+SETUPFILE: ../../aurellem/org/setup.org
rlm@22 7 #+INCLUDE: ../../aurellem/org/level-0.org
rlm@22 8 #+babel: :mkdirp yes :noweb yes
rlm@22 9
rlm@22 10 * Background
rlm@22 11 Artificial Intelligence has tried and failed for more than half a
rlm@22 12 century to produce programs as flexible, creative, and "intelligent"
rlm@22 13 as the human mind itself. Clearly, we are still missing some important
rlm@22 14 ideas concerning intelligent programs or we would have strong AI
rlm@22 15 already. What idea could be missing?
rlm@22 16
rlm@22 17 When Turing first proposed his famous "Turing Test" in the
rlm@34 18 groundbreaking paper [[../sources/turing.pdf][/Computing Machines and Intelligence/]], he gave
rlm@22 19 little importance to how a computer program might interact with the
rlm@22 20 world:
rlm@22 21
rlm@22 22 #+BEGIN_QUOTE
rlm@22 23 \ldquo{}We need not be too concerned about the legs, eyes, etc. The example of
rlm@22 24 Miss Helen Keller shows that education can take place provided that
rlm@22 25 communication in both directions between teacher and pupil can take
rlm@22 26 place by some means or other.\rdquo{}
rlm@22 27 #+END_QUOTE
rlm@22 28
rlm@22 29 And from the example of Hellen Keller he went on to assume that the
rlm@22 30 only thing a fledgling AI program could need by way of communication
rlm@22 31 is a teletypewriter. But Hellen Keller did possess vision and hearing
rlm@22 32 for the first few months of her life, and her tactile sense was far
rlm@22 33 more rich than any text-stream could hope to achieve. She possessed a
rlm@22 34 body she could move freely, and had continual access to the real world
rlm@22 35 to learn from her actions.
rlm@22 36
rlm@22 37 I believe that our programs are suffering from too little sensory
rlm@22 38 input to become really intelligent. Imagine for a moment that you
rlm@22 39 lived in a world completely cut off form all sensory stimulation. You
rlm@22 40 have no eyes to see, no ears to hear, no mouth to speak. No body, no
rlm@22 41 taste, no feeling whatsoever. The only sense you get at all is a
rlm@22 42 single point of light, flickering on and off in the void. If this was
rlm@22 43 your life from birth, you would never learn anything, and could never
rlm@22 44 become intelligent. Actual humans placed in sensory deprivation
rlm@22 45 chambers experience hallucinations and can begin to loose their sense
rlm@34 46 of reality. Most of the time, the programs we write are in exactly
rlm@34 47 this situation. They do not interface with cameras and microphones,
rlm@34 48 and they do not control a real or simulated body or interact with any
rlm@34 49 sort of world.
rlm@22 50
rlm@22 51 * Simulation vs. Reality
rlm@22 52 I want demonstrate that multiple senses are what enable
rlm@22 53 intelligence. There are two ways of playing around with senses and
rlm@22 54 computer programs:
rlm@22 55
rlm@34 56
rlm@34 57 ** Simulation
rlm@22 58 The first is to go entirely with simulation: virtual world, virtual
rlm@22 59 character, virtual senses. The advantages are that when everything is
rlm@22 60 a simulation, experiments in that simulation are absolutely
rlm@22 61 reproducible. It's also easier to change the character and world to
rlm@22 62 explore new situations and different sensory combinations.
rlm@22 63
rlm@22 64 If the world is to be simulated on a computer, then not only do you
rlm@22 65 have to worry about whether the character's senses are rich enough to
rlm@22 66 learn from the world, but whether the world itself is rendered with
rlm@22 67 enough detail and realism to give enough working material to the
rlm@22 68 character's senses. To name just a few difficulties facing modern
rlm@22 69 physics simulators: destructibility of the environment, simulation of
rlm@22 70 water/other fluids, large areas, nonrigid bodies, lots of objects,
rlm@22 71 smoke. I don't know of any computer simulation that would allow a
rlm@22 72 character to take a rock and grind it into fine dust, then use that
rlm@22 73 dust to make a clay sculpture, at least not without spending years
rlm@22 74 calculating the interactions of every single small grain of
rlm@22 75 dust. Maybe a simulated world with today's limitations doesn't provide
rlm@22 76 enough richness for real intelligence to evolve.
rlm@22 77
rlm@34 78 ** Reality
rlm@22 79
rlm@22 80 The other approach for playing with senses is to hook your software up
rlm@22 81 to real cameras, microphones, robots, etc., and let it loose in the
rlm@22 82 real world. This has the advantage of eliminating concerns about
rlm@22 83 simulating the world at the expense of increasing the complexity of
rlm@22 84 implementing the senses. Instead of just grabbing the current rendered
rlm@22 85 frame for processing, you have to use an actual camera with real
rlm@22 86 lenses and interact with photons to get an image. It is much harder to
rlm@22 87 change the character, which is now partly a physical robot of some
rlm@22 88 sort, since doing so involves changing things around in the real world
rlm@22 89 instead of modifying lines of code. While the real world is very rich
rlm@22 90 and definitely provides enough stimulation for intelligence to develop
rlm@22 91 as evidenced by our own existence, it is also uncontrollable in the
rlm@22 92 sense that a particular situation cannot be recreated perfectly or
rlm@22 93 saved for later use. It is harder to conduct science because it is
rlm@22 94 harder to repeat an experiment. The worst thing about using the real
rlm@22 95 world instead of a simulation is the matter of time. Instead of
rlm@22 96 simulated time you get the constant and unstoppable flow of real
rlm@22 97 time. This severely limits the sorts of software you can use to
rlm@22 98 program the AI because all sense inputs must be handled in real
rlm@22 99 time. Complicated ideas may have to be implemented in hardware or may
rlm@22 100 simply be impossible given the current speed of our
rlm@22 101 processors. Contrast this with a simulation, in which the flow of time
rlm@22 102 in the simulated world can be slowed down to accommodate the
rlm@22 103 limitations of the character's programming. In terms of cost, doing
rlm@22 104 everything in software is far cheaper than building custom real-time
rlm@22 105 hardware. All you need is a laptop and some patience.
rlm@22 106
rlm@22 107 * Choose a Simulation Engine
rlm@22 108
rlm@22 109 Mainly because of issues with controlling the flow of time, I chose to
rlm@34 110 simulate both the world and the character. I set out to make a world
rlm@34 111 in which I could embed a character with multiple senses. My main goal
rlm@34 112 is to make an environment where I can perform further experiments in
rlm@34 113 simulated senses.
rlm@22 114
rlm@34 115 I examined many different 3D environments to try and find something I
rlm@34 116 would use as the base for my simulation; eventually the choice came
rlm@34 117 down to three engines: the Quake II engine, the Source Engine, and
rlm@34 118 jMonkeyEngine.
rlm@22 119
rlm@27 120 ** [[http://www.idsoftware.com][Quake II]]/[[http://www.bytonic.de/html/jake2.html][Jake2]]
rlm@22 121
rlm@22 122 I spent a bit more than a month working with the Quake II Engine from
rlm@22 123 ID software to see if I could use it for my purposes. All the source
rlm@22 124 code was released by ID software into the Public Domain several years
rlm@22 125 ago, and as a result it has been ported and modified for many
rlm@22 126 different reasons. This engine was famous for its advanced use of
rlm@22 127 realistic shading and had decent and fast physics
rlm@197 128 simulation. Researchers at Princeton [[http://papers.cnl.salk.edu/PDFs/Intracelllular%20Dynamics%20of%20Virtual%20Place%20Cells%202011-4178.pdf][used this code]] ([[http://brainwindows.wordpress.com/2009/10/14/playing-quake-with-a-real-mouse/][video]]) to study spatial
rlm@22 129 information encoding in the hippocampal cells of rats. Those
rlm@22 130 researchers created a special Quake II level that simulated a maze,
rlm@22 131 and added an interface where a mouse could run around inside a ball in
rlm@22 132 various directions to move the character in the simulated maze. They
rlm@22 133 measured hippocampal activity during this exercise to try and tease
rlm@22 134 out the method in which spatial data was stored in that area of the
rlm@22 135 brain. I find this promising because if a real living rat can interact
rlm@22 136 with a computer simulation of a maze in the same way as it interacts
rlm@22 137 with a real-world maze, then maybe that simulation is close enough to
rlm@22 138 reality that a simulated sense of vision and motor control interacting
rlm@22 139 with that simulation could reveal useful information about the real
rlm@24 140 thing. There is a Java port of the original C source code called
rlm@24 141 Jake2. The port demonstrates Java's OpenGL bindings and runs anywhere
rlm@24 142 from 90% to 105% as fast as the C version. After reviewing much of the
rlm@24 143 source of Jake2, I eventually rejected it because the engine is too
rlm@24 144 tied to the concept of a first-person shooter game. One of the
rlm@24 145 problems I had was that there do not seem to be any easy way to attach
rlm@24 146 multiple cameras to a single character. There are also several physics
rlm@24 147 clipping issues that are corrected in a way that only applies to the
rlm@24 148 main character and does not apply to arbitrary objects. While there is
rlm@24 149 a large community of level modders, I couldn't find a community to
rlm@24 150 support using the engine to make new things.
rlm@22 151
rlm@27 152 ** [[http://source.valvesoftware.com/][Source Engine]]
rlm@22 153
rlm@22 154 The Source Engine evolved from the Quake II and Quake I engines and is
rlm@22 155 used by Valve in the Half-Life series of games. The physics simulation
rlm@22 156 in the Source Engine is quite accurate and probably the best out of
rlm@22 157 all the engines I investigated. There is also an extensive community
rlm@22 158 actively working with the engine. However, applications that use the
rlm@22 159 Source Engine must be written in C++, the code is not open, it only
rlm@22 160 runs on Windows, and the tools that come with the SDK to handle models
rlm@22 161 and textures are complicated and awkward to use.
rlm@22 162
rlm@27 163 ** [[http://jmonkeyengine.com/][jMonkeyEngine3]]
rlm@22 164
rlm@22 165 jMonkeyEngine is a new library for creating games in Java. It uses
rlm@22 166 OpenGL to render to the screen and uses screengraphs to avoid drawing
rlm@22 167 things that do not appear on the screen. It has an active community
rlm@22 168 and several games in the pipeline. The engine was not built to serve
rlm@22 169 any particular game but is instead meant to be used for any 3D
rlm@22 170 game. After experimenting with each of these three engines and a few
rlm@22 171 others for about 2 months I settled on jMonkeyEngine. I chose it
rlm@22 172 because it had the most features out of all the open projects I looked
rlm@22 173 at, and because I could then write my code in Clojure, an
rlm@34 174 implementation of LISP that runs on the JVM.
rlm@29 175
rlm@29 176
rlm@29 177
rlm@29 178
rlm@29 179
rlm@29 180
rlm@29 181
rlm@29 182
rlm@29 183
rlm@29 184
rlm@29 185
rlm@29 186
rlm@29 187
rlm@29 188
rlm@29 189
rlm@29 190
rlm@29 191
rlm@29 192
rlm@29 193
rlm@29 194
rlm@29 195
rlm@29 196
rlm@29 197
rlm@29 198
rlm@29 199
rlm@29 200
rlm@29 201
rlm@29 202
rlm@29 203
rlm@29 204
rlm@29 205
rlm@32 206
rlm@32 207