Invalidate a Theory
- One important scientific task is to invalidate theories.
- People spend time depending on a theory.
- If that theory is wrong, and you show it's wrong, you save
a lot of time.
- An example of this is Minsky and Papert's Proof that Perceptrons
can't learn exclusive Or.
- This led to problems with funding for neural network research
for a decade, but really showed a problem.
- One of the problems with this is that if you can't actually prove it,
you just become part of the debate (which is not that much
of a problem).
- Minsky and Papert were already famous, but this would have made
my career.
- You can also propose new experiments that at least could
invalidate a theory.
- Thought experiments can be useful with this (and with other tasks).
