It is clear that humans would need to be involved in this global brain, and as such there would be several human limitations. I am not sure what kinds of hierarchical prediction-structures with human components would be effective. It is possible to view cells as making simple economic choices I think, but am not sure how hierarchical structures in the brain - or any organic structures - are produced at genesis, so I have no good inspiration here!

I know that hierarchical structures are particularly valuable because they maximize the ability to perceive distinction while preserving unity. The key to perceived beauty is the apparent complexity combined with underlying unity. We need to be able to differentiate a billion different sentences or pictures, while having the ability to integrate each particular sentence or picture into a unitary whole.

First hit.

I love posts like this!

Perceptrons or Why Does the Brain Predict?

The brain is mainly a prediction-machine. It takes sensory input and through the process of perception figures out what sensory input will occur next. It does this by compressing sensory input in both time and space; if you see an occluded figure, you'll typically still recognize it. If you hear part of a song, you'll typically still recognize it. The brain has compressed the total figure and song, and will be able to predict its presence based on any fraction of the total information.

Why does the brain predict? To make useful actions possible. All moving creatures need to predict the consequences of both sensory input and the result of their own actions (output). This is why brains exist - to coordinate perception with action. At first, nervous systems were used to react immediately to stimuli, but over time the reactions grew to become thoughtful. At heart, the ability to predict the affordances of stimuli and the consequences of action are fundamental.

How do our run-of-the-mill IQ tests try to assess our intelligence? "Which number comes next?" "Pick the figure that completes the pattern," etc. Here we typically expect someone to complete a pattern on the basis of some limited sensory experience by means of finding an elegant way of compressing the given input where the unknown comes out simply.

There are also lots of patterns that are irrational in the sense that they cannot be compressed elegantly to reveal the unknown. To predict events in the world we typically mix these two forces: On the one hand we look for elegant, beautiful ways to compress the total information we are given. On the other hand, we are collecting a vast store of known patterns that exist in the world even if they are ugly (not easily compressible).

What can we draw from all this? First, prediction-markets should be seen as trying to complete incomplete knowledge in two senses, rationally and empirically, and never just temporally. Second, we should recognize that "incomplete knowledge" is a subjective feature that apply to limited perspectives, whether the limit be in intellectual power e.g. we can't compute the answer, scientific insight e.g. we don't know if it's empirically possible, or empirical knowledge e.g. we don't know if a particular fact is true.

I also want to mention that Perceptrons can take hypotheticals as input; IF x, then ?. This is what happens when Perceptrons are used to predict the consequences of the predictors action, it puts the given action as a hypothetical x, and then tries to predict what will happen. In this way we get the additional result that prediction-market also provides predictions about hypothetical situations.

1) Rapid loss of value of 50% in hours is very rare, even for bitcoin (once the network is established)... can anyone show an example where Bitcoin lost 50% in hours (excluding the Mt. Gox flash crash)?

I'm still crying for lack of buyorder @BTC-e today.

The holder of BitUSD is not protected against systemic risks associated with all crypoto-currencies, only from the volatility that is independent of those systemic risks.   

+3%

Excellent points cloudcoiner!

This is quickly becoming the coolest DAC on the block. 

Interesting.