rlm@302: Dear Professor Winston,
rlm@302: 
rlm@325: I've finished the first part of my project, building a
rlm@325: framework for virtual sensate creatures; I would like your
rlm@325: help evaluating what I've done so far, and deciding what to
rlm@325: do next.
rlm@302: 
rlm@325: For the work I've done so far, I compiled the results into
rlm@325: short articles that explain how I implemented each sense,
rlm@325: with videos that show each sense in action. Please look
rlm@325: through the articles, in particular the video showcase, and
rlm@325: tell me what you think.
rlm@302: 
rlm@310: Video Showcase : http://aurellem.org/cortex/org/cover.html
rlm@302: 
rlm@302: Introduction:
rlm@302: http://aurellem.org/cortex/html/intro.html
rlm@302: http://aurellem.org/cortex/html/sense.html
rlm@302: 
rlm@305: Physical Bodies : http://aurellem.org/cortex/html/body.html
rlm@305: Vision  : http://aurellem.org/cortex/html/vision.html
rlm@305: Hearing : http://aurellem.org/cortex/html/hearing.html
rlm@305: Touch   : http://aurellem.org/cortex/html/touch.html
rlm@325: Proprioception : http://aurellem.org/cortex/html/proprioception.html
rlm@305: Muscles            : http://aurellem.org/cortex/html/movement.html
rlm@305: Full Demonstration : http://aurellem.org/cortex/html/integration.html
rlm@302: 
rlm@325: I think this work could be a fruitful foundation for a
rlm@325: Master's thesis, so in particular, I'd like critiques,
rlm@325: suggestions, and project ideas. For example, here are some
rlm@325: projects I think would be worthy, in increasing order of
rlm@325: complexity:
rlm@302: 
rlm@325:     * Create a self-powered joint that can determine its
rlm@325:       range of motion and joint type (hinge, cone,
rlm@325:       point-to-point, etc.) by making exploratory muscle
rlm@325:       movements and observing their effect.
rlm@302: 
rlm@325:     * Develop an agent that writes and debugs low-level
rlm@325:       motor control programs to achieve simple goals like
rlm@325:       "look at the light" or "extend all of your
rlm@325:       fingers". These simple "calisthenic" programs could
rlm@325:       then be combined to form more elaborate procedures of
rlm@325:       motion, which in turn could be the basic instinctive
rlm@325:       reflexes in the "spinal cord" of some more advanced
rlm@325:       creature. (like Sussman's HACKER program but in a
rlm@325:       richer world)
rlm@302: 
rlm@305:     * Program a group of creatures that cooperate with each
rlm@325:       other. Because the creatures would be simulated, I
rlm@325:       could investigate computationally complex rules of
rlm@325:       behavior which still, from the group's point of view,
rlm@325:       would happen in "real time". Interactions could be as
rlm@325:       simple as cellular organisms communicating via
rlm@325:       flashing lights, or as complex as humanoids completing
rlm@325:       social tasks, etc.
rlm@302: 
rlm@325:     * Simulated Imagination -- this would involve a creature
rlm@325:       with an effector which creates an entire new
rlm@325:       sub-simulation where the creature has direct control
rlm@325:       over placement/creation of objects via simulated
rlm@325:       telekinesis. The creature observes this sub-world
rlm@325:       through it's normal senses and uses its observations
rlm@325:       to make predictions about its top level world.
rlm@302: 
rlm@325:     * Integrate the simulated world with Genesis, so that
rlm@325:       Genesis could use the simulated world to answer
rlm@325:       questions about a proposed physical scenario. For
rlm@325:       example "You stack two blocks together, then hit the
rlm@325:       bottom block with your hand. Does the top block
rlm@325:       move?". This project is complicated and very large in
rlm@325:       scope, but it could be narrowed to focus on a single
rlm@325:       key aspect. For example, one key aspect of turning a
rlm@325:       scenario into a simulation is knowing when you're
rlm@325:       constructing "typical" or "atypical" examples of the
rlm@325:       scenario. So, a narrower project might simply learn
rlm@325:       about the edge cases of different scenarios (e.g. "A
rlm@325:       block stacked on top of another block is usually
rlm@325:       stable, provided the bottom block is large enough, and
rlm@325:       is not moving, and is level, etc."). With this
rlm@325:       knowledge, this kind of program could aid Genesis not
rlm@325:       only in answering common-sense questions, but in
rlm@325:       refining them: "A block is stacked on top of another
rlm@325:       block. Is it stable?"; "Usually, but do you know if
rlm@325:       the bottom block is slanted?", etc.
rlm@302: 
rlm@325: These are some ideas, but I think you can come up with
rlm@325: better ones. I can't wait to hear your critiques and
rlm@325: suggestions.
rlm@302: 
rlm@325: Finally, regarding next year at MIT, can I be considered for
rlm@325: the position of TA for 6.034 or 6.xxx? Also, do you want me
rlm@325: to return to MIT at the beginning of Fall or at the
rlm@325: beginning of Summer?
rlm@302: 
rlm@305: Sincerely, 
rlm@302: --Robert McIntyre