April 24, 2015

Big questions

While browsing some interesting talks on the web, I realized some of the "big questions" that cause much confusion fall into one of two categories.
  1. Physics does not have X. Is X something real and missing in physics, or is X an illusion, a modeling tool, a result of our limited information, agent based perspective, or stupidity? Examples: consciousness, free will, causality and directionality of time.
  2. Physics does have X which doesn't make sense. Is X something real, or a modeling tool? Examples: randomness, entropy, various sorts of quantum weirdness.
Here are some of the related links:
  • I was trying to find out what fellow MIT AI Lab alum Gary Drescher is up to after writing Good and Real (which covers most of the big questions mentioned above). I found this video of a talk on causality and choice he gave at the Singularity Summit in 2009. He argues convincingly that determinism and choice are not at odds with each other.
  • Speaking of causality, Michael Nielsen (author of Quantum Computation and Quantum Information, as well as a free online book on Neural Networks and Deep Learning) has a nice article summarizing the key points of Judea Pearl's Causality book. The epilogue of the book (well worth the read if you don't have the time or patience for the whole book) is a lecture Pearl gave some years ago where he covers the history of the confusion about causality, why physics (if we model the whole universe) does not require it, how economics and other social sciences (where lack of controlled experiments make causality detection difficult) still do not adequately model it, and how it is actually possible in some cases to derive causality from purely observational data without resorting to controlled experiments.
  • Another regular at the Singularity Summit is Eliezer Yudkowsky. I first discovered his posts on Less Wrong, which has a lot of good material that identifies common sources of confusion when thinking about the big questions.


Unknown said...

Isn't the simulation (Newcomb's dilemma) an example of infinite regress? The machine is not simulating an external environment, it's simulating an environment that it's part of. Then it has to simulate itself that simulates itself that simulates itself ...

Is such a simulation even possible?

A quick google shows arguments for infinite regress, but only for prison dilemma.

Unknown said...

while reading https://www.goodreads.com/book/show/59695048-existential-physics, I remembered my decade old comment on this thread. Deutsch thinks that it's not possible, for the same reason. I felt a tiny bit proud that I had similar thoughts with such a giant. He named his ideas under "constructor theory": https://en.wikipedia.org/wiki/Constructor_theory.