thread

Dear Jeffrey,

It was very fun talking with you the weekend before last. I have downloaded The Improvement of the Mind and am reading. I feel my mind improving already ;-)

The conversation and your ideas fit in closely with work I've been doing and am proposing to do. Maybe this is not surprising because we've been talking about these topics for years. I've been traveling around (Santa Fe, now in San Francisco, soon to be in Banff) and so have had time to think more about what we discussed.

Here is a succinct summary of my own take. I know yours is somewhat different.

Information is a fundamental quantity, measured in bits.

Information can be random, like the typical string of bits one gets by flipping a coin 010111011010100001 (I just flipped a coin and let heads = 1 and tails = 0), or it can be ordered, like the bit string 0000000000000000.

There is a technical definition of order and randomness: a bit string is ordered if there is a succinctly describable method, e.g., a short computer program, for producing it. By contrast, a string is random if the shortest program for producing it is the same length as the string itself. For example, the string consisting of a billion 0's can be produced by a short program: Print '0' 10^9 times. By contrast, the shortest program to produce the string 010111011010100001 is something like: Print 010111011010100001. This way of defining order/randomness is called algorithmic information.

The interesting thing about algorithmic information is that the short program can be hard to find. A string can look very random and still have a short program. For example, the first billion bits of pi, written in binary, have a short program, but if I just give you those bits, they would look statistically random.

This means that so far as we or any other information processing system is concerned, order/randomness is subjective: some information can be ordered and non-random, but we may not be able to recognize the underlying order, so we treat it as random. That is, the order is cryptic: we don't know how to decode it. This crypticity is the key feature for looking at both physical systems and human/social systems.

Physical systems:

In physical systems such as a gas of molecules, entropy is the amount of information that is required to describe the underlying motions of those molecules. Because the molecules are bouncing off each other in an essentially random way, entropy is assumed to be random information.

Heat is energy that is encumbered by entropy/randomness. The randomness makes it harder to take advantage of that energy.

By contrast, free energy is energy where the information required to describe how that energy is arranged is ordered, not random. Free energy is energy we can take advantage of.

Now comes the kicker: suppose that a system is actually ordered, but that order is cryptic. If we can't decipher the order, we can't take advantage of it.

So whether energy is free or not depends on our ability to detect its underlying order. That is, the availability of energy depends on the computational/decrypting ability of whatever system (molecule, microbe, human) is trying to take advantage of that energy.

Social/human interactions:

Human interactions are about the exchange of information, plus other stuff (goods, services, money, etc.). Everything that is exchanged brings with it the information that describes what is exchanged, what can be done with it, etc. So for example, a US Treasury bond comes with the specification of its price and its future interest payments. A complex option comes with the specification of what can be bought and sold when.

As a result, human interactions are awash in information. Different people are capable of decoding/deciphering that information in different ways. The ability to detect a pattern or order in information translates into the ability to take advantage of a social situation. For example, the efficient market hypothesis states that fluctuations in the prices of a given stock should be essentially random. But if you happen to possess some information that allows you to predict the future fluctuations of the stock, then you can make money.

Similarly, to play into your ideas about deception, when two actors enter into a transaction, each presents the other with information about their obligations under the transaction. Each actor looks at that information and judges whether the transaction will turn out to their own advantage. But because each actor perceives different patterns in the information, they can come up with different evaluations of the future worth of the transaction.

Deception arises because one actor may hide a pattern in the revealed information, a pattern that the other actor doesn't perceive, but that makes the transaction more advantageous to the first actor. I've been told that this is called, doing business.

Talk some more?

Yours,
Seth

1, re flips. ordering is not my focus. my focus is that the program that says. as you approach a larger number the total number of ones and zeros ( the fair coin flip ), should be 50/50/ it is the opposite of information as that relates to the individual flips. , I maintain that the distribution of flips may be described by skewing. A force that leads to 50/ 50

2. again re heat. we cant say anything about indiv particles. but we can measure their overall temp and pressure

3.40 social distributions of talents and characteristics, seem to follow the same distribution for years.

4, deception, relies on the concept of INTENTION. not well described.

1. Yes this is 'reversion to the mean' and it definitely happens. This is the basis for what is called anentropic force

(rubber bands as we discussed)

2. Absolutely: for large numbers of particles, the entropy arguments give rise to macroscopic

statistical quantities like temperature and pressure

3. Yes: and they don't obey the central limit theorem (they are non-Gaussian). As we discussed,

there are lots of potential reasons for this. The main one is that if there is no restriction on

the variance of a distribution the central limit theorem doesn't apply and so you get power laws

as in the Pareto law for income distribution.

4. I didn't claim I described it well. I'd say one actor intentionally hides some information that the

other actor can't decrypt. That is, the first actor is trying to deceive the second in order

to take advantage.

Got to run give a talk for Stewart Brand/Danny Hillis! Seth