Why You Should Be Eager to Be Wrong
Rich Hua stashed this in Talent Code
"Uncertainty is a very mentally demanding, and in a certain way, physically demanding process."
Loved this article. As F. Scott Fitzgerald said, "The test of a first-rate intelligence is the ability to hold two opposed ideas in mind at the same time and still retain the ability to function." I feel like Bertrand Russell said something similar but can't place it right now.
One way to improve one's facility with uncertain/conflicting evidence is to study the basics of fuzzy logic. The basic math is simple -- accessible to almost anyone with a little effort -- and beautiful. In a nutshell, truth values can be 0 (false), 1, (true), or anywhere in between. The middle values can represent probabilities, but than can also represent partial attainment.
For example, am I close to your house? We can answer that with a fuzzy value... 0 when I haven't left, 1 when I'm at your house, and increasing gradually from 0 to 1 as I approach.
The coolest thing to me is how fuzzy logic extends, rather than conflicts with, classical Boolean logic. For example the fuzzy OR function is just the minimum of the 2 truth values. But this same function (min) gives same results as a classical Boolean truth table (e.g. 0 OR 1 = 0, which is the minimum of 0 and 1)
Fuzzy if-thens can be very powerful, and have been the basis for relatively simple software solutions that can be very difficult to implement with classical logic.
Bart Kosko has a pop science intro book on this, which is good. And I'm sure there are some good introductions online too.
That's a pretty fascinating model.
Here I thought machines were nothing but zeroes and ones.
There's actually a lot more to it than that.
So now I'm beginning to believe that AI really can make good decisions.
Is that a bad leap?
I guess it depends on how you define AI... I don't think we're in any real danger of it soon, but this fuzzy model has been huge for problems under that heading. I believe that it was fuzzy control systems that actually turned the tide on digital appliances for example... could be wrong about that. Here's an example of a fuzzy control system based on fuzzy if-then rules... you can see how intuitive this would be versus "binary" thinking about when to alter various system parameters. The model can be set up by a human and then the system can be trained to find the right "weights" (fuzzy values), without anyone ever having to work with the raw numbers.
This seems like a much better model of how humans often think about things, i.e. rather than "all or nothing" (Boolean/binary) we think things to various degrees. Maybe inside our brains, the thresholds for neuron firing *could* be quantified, but we never experience it this way... and yet, the system works even we we never know the specific quantities involved.
It does seem like a better model than human typically use.
Maybe you're right that this is how our internals work but they only manifest themselves through intuition.