For a more realistic example, this phenomenon of a more accurate model being worse is a common issue in database query optimization.
When a user runs a SQL query, the optimizer uses statistics about the data to estimate the cost of many different ways to execute the query, then picks the plan with the cheapest estimate.
When the optimizer misestimates the cost and chooses a bad plan, the typical solution is to add more detailed statistics about the data. But occasionally adding more statistics can cause the optimizer to choose a plan that's actually worse.
For your second question, this paper describes the square root law, though in a somewhat different setting: Strong profiling is not mathematically optimal for discovering rare malfeasors | PNAS. (Incidentally, a friend of mine used this once in an argument against stop-and-frisk.)
It doesn't give a complete proof, though it describes it as a "straightforward minimization with a Lagrange multiplier".
This distinction reminds me of the battles in Ender's Game.
As I recall, Ender was the overall commander, but he delegated control of different parts of the fleet to various other people, as most modern militaries do.
The bugs fought as a hive mind, and responded almost instantly across the entire battlefield, which made it challenging for the humans to keep up in large-scale battles.
You might be interested in reading this paper, which tries to build a vision for just such a tool: Towards a Dynamic Multiscale Personal Information Space (mit.edu).
As far as I know, nothing like it exists yet.
As I understand it you've split this idea into two parts:1. Good - People can develop their intuition to make correct predictions even without fully understanding how they can tell.2. Bad - People frequently make mistakes when they jump to conclusions based on subconscious assumptions.
I'm confident that (2) is true, but I'm partly skeptical of (1). For example, Scott reviewed one of John Gottman's books, and concluded that it's "totally false" that he can predict who will get divorced 90% of the time: Book Review: The Seven Principles For Making Marriage Work | Slate Star Codex.
Another example which you've probably heard before is from the Superforecasting book: Book Review: Superforecasting | Slate Star Codex. Tetlock discovered that most pundit predictions were worse than random in the domains in which they were supposed to be experts.
I believe that intuition works well in "kind" learning environments, but not in "wicked" environments. David Epstein distinguished these in his book, Range: Why Generalists Triumph in a Specialized world (davidepstein.com). The main difference between the two environments is whether there's quick and reliable feedback on whether your judgment was correct.
Chess is a kind environment, which is why Magnus Carlsen can play dozens of people simultaneously and still win virtually every game. This also explains the results of the Bechara gambling task. Politics and marriage are wicked learning environments, which is why expert predictions are much less reliable.
Did you really mean 6 AM? I was thinking of joining, but I don't usually go anywhere on Saturdays before 10. :P
I understood this sentence to mean that animals are intelligent insofar as they can pick up subtle body language cues.
There's a few one-star Amazon reviews for the book that suggest McAfee's data is incorrect or misleading. Here's a quote from one of them, which seems like a solid counterargument to me:
"However, on the first slide on page 79, he notes that the data excludes impact from Import/export of finished goods. Not raw materials but finished goods. He comments that Net import is only 4% of GDP in the US. Here he makes a (potentially) devastating error – (potentially) invalidating his conclusion.While Net imports is indeed around 4% of GDP, the gross numbers are Exports at approx. +13% and Imports at approx. -17%. So any mix difference in finished goods in Export and Import, can significantly change the conclusion. It so happens that US is a major Net importer of finished goods e.g. Machinery, electronic equipment and autos (finished goods, with materials not included above in the consumption data). Basically, a big part of US’ consumption of cars, washing machines, computers etc. are made in Mexico, China etc. They contain a lot of materials, not included in the graphs, upon which he builds his conclusion/thesis. So quite possibly, there is no de-coupling."
Thanks for the review - I appreciate that you spent the time to sift through these ideas.
When you say that we need to "be willing to make uncomfortable, difficult changes," do you have some specific changes in mind?
That depends what you mean by "computational problems". In its usual definition, a Turing machine takes a single input string, and provides a single bit as output: "Accept" or "Reject". (It may also of course run forever without providing an answer.)
For example, the question, "What is the shortest path through this list of cities?" isn't possible to encode directly into a standard Turing machine. Instead, the Turing machine can answer, "Is there a path through this list of cities with length at most k?", for some constant k.
If you don't like this, there are ways to modify the definition of a Turing machine. But for the purposes of studying computational complexity, all reasonable definitions seem to be equivalent.