Brain Signals Shown to Move a Robot's Arm
http://www.nytimes.com/2000/11/16/science/16ROBO.html
[From Rick Marken (2000.11.16.0900)]
Norman T. Hovda (2000.11.16) --
Brain Signals Shown to Move a Robot's Arm
http://www.nytimes.com/2000/11/16/science/16ROBO.html
I think these two quotes from the article are very telling:
Dr. Nicolelis offered this example: If someone moves to push open
a heavy door, the brain generates signals telling the arms and
legs exactly how much pressure to expect and how to maintain
balance. The plan, based on previous experiences with heavy doors,
is made a half-second before commands are sent down the spinal cord
and out to muscles and joints where the movement is carried out,
he said.
Using advanced computational techniques, the researchers identified
patterns used to plan the reaching movements and transformed those
patterns into numerical instructions that could operate a robot.
Apparently these researchers are committed to the idea that
efferent brain signals are specifications for output (the
S-R view) rather than input (the PCT view). Too bad.
Best
Rick
···
--
Richard S. Marken Phone or Fax: 310 474-0313
MindReadings.com mailto: marken@mindreadings.com
www.mindreadings.com
[From Bruce Nevin (2000.11.17.0719 EST)]
Rick Marken (2000.11.16.0900)--
Norman T. Hovda (2000.11.16) --
> Brain Signals Shown to Move a Robot's Arm
> http://www.nytimes.com/2000/11/16/science/16ROBO.htmlApparently these researchers are committed to the idea that
efferent brain signals are specifications for output (the
S-R view) rather than input (the PCT view). Too bad.
When I heard about this on NPR they specifically mentioned feedback for e.g. grasping an egg without crushing it, so they're more aware than this excerpt would indicate. Probably they're thinking of adding bits of feedback here and there as "enhancements" to the CogSci "planning" view of things. They had to have a large bank of PCs in parallel to run the arm. Most of the computation, they said, went into filtering out most of the neural signals they were picking up and keeping only those that worked as muscle commands.
What they're not getting is that the "plan" that issues the muscle intensity "commands" is made up on the fly at a higher level in the hierarchy, and the "plan" that issues configuration "commands" to that level is made up on the fly at a higher level ... all as a function of continuous negative feedback control. Sounds very close.
Bruce Nevin
···
At 08:56 AM 11/16/2000 -0800, Richard S. Marken wrote:
[From Rick Marken (2000.11.17.0730)]
Me:
Apparently these researchers are committed to the idea that
efferent brain signals are specifications for output (the
S-R view) rather than input (the PCT view). Too bad.
Bruce Nevin (2000.11.17.0719 EST)
When I heard about this on NPR they specifically mentioned
feedback for e.g. grasping an egg without crushing it, so
they're more aware than this excerpt would indicate.
Yes. All psychologists are aware that "feedback" is involved
in behavior. They just don't know _how_ it's involved. They
seem to think that feedback (in the form of perceptions of
the results of action) _guides_ action. What they don't get
is the fact that, in a control loop, action actually _guides_
feedback (perception), keeping it in a specified (reference)
state.
What they're not getting is that the "plan" that issues the
muscle intensity "commands" is made up on the fly at a higher
level in the hierarchy
I think they get this just fine. What they don't get is the
fact that a "plan" issues "commands" for _perception_
(of variables like muscle intensity), not for output (of those
same variables).
Sounds very close [to PCT].
Not to me.
Best
Rick
···
---
Richard S. Marken Phone or Fax: 310 474-0313
MindReadings.com mailto: marken@mindreadings.com
www.mindreadings.com