Connectionists: The meaning of "meaning"

[Rothganger, Fred]

There have been several interesting threads on this list recently about "meaning". I'd like to suggest a way of thinking about this, based on who/what the observer is:

  1.
Another agent (animal or human) -- One agent performs a physical action that stimulates the brain of the other. It is useful, but not necessary, that actions can be quantized into a discrete set called "symbols". It's also helpful for both agents to have some rough agreement on the referent for a given symbol.
  2.
A neuroscientist -- The experimenter observes that some neurons fire more actively when a specific stimulus is presented to the brain, and concludes that the neurons "represent" the stimulus.
  3.
The brain itself -- This is similar to case #1, but for internal consumption. Neural activity could potentially, but not necessarily, form discrete sets that allow operations called "computation". It's also possible that neural activity is more continuous-valued and involved in feedback control or other processes that don't neatly fall into symbolic computation.

All three of these are valid points of view. There is a danger, though, in confusing them with each other. For example, confusing #2 with #3. In that case the neuroscientist accidentally creates a homunculus by inserting themself into brain's processing and believing that it uses the neural activity the same way the neuroscientist does. There are numerous ways in which this can lead to a wrong conclusion.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20240618/4d8f4e69/attachment.html>