http://motionmountain.org/ seems a good source, 50 mb pdf
random memory.. "Obviously every record is an object.But under which conditions does an object qualify as a record?A signature can be the record of the agreement on a commercial transaction. A single small dot of ink is not a record,because it could have appeared by mistake,for example by an accidental blot.In contrast,it is improbable that ink should fall on paper exactly in the shape of a signature.(e simple signatures of physicians are obviously exceptions.)Simply speaking,a record is any object,which,in order to be copied,has to be forged.More precisely,a record is an object or a situation that cannot arise nor disappear by mistake or by chance.Our personal memories,be they images or voices, have the same property;we can usually trust them,because they are so detailed that they cannot have arisen by chance or by uncontrolled processes in our brain. Can we estimate the probability for a record to appear or disappear by chance?Yes,we can.Every record is made of a characteristic number N of small entities,for example the number of the possible ink dots on paper,the number of iron crystals in a cassette tape, the electrons in a bit of computer memory,the silver iodide grains in a photographic negative,etc.e chance disturbances in any memory are due to internal uctuations, also called noise .Noise makes the record unreadable;it can be dirt on a signature,thermal magnetization changes in iron crystals,electromagnetic noise inside a solid state memory, etc.Noise is found in all classi ers,since it is inherent in all interactions and thus in all information processing. It is a general property that internal uctuations due to noise decrease when the size, i.e.the number of components of the record is increased.In fact,the probability p mis for a misreading or miswriting of a record changes as p mis N ,(464) where N is the number of particles or subsystems used for storing it.is relation appears because,for large numbers,the so called normal distribution is a good approximation of almost any process;the width of the normal distribution,which determines the probab ility of record errors,grows less rapidly than its integral when the number of entities is increased.(Are you able to con rm this?)Challenge 1144 ny We conclude that any good record must be made from a large number of entities.e larger the number,the less sensitive the memory is to uctuations.Now,a systemof large size with small uctuations is called a (physical)bath .Only baths make memories possible. In other words,every record contains a bath.We conclude that any observation of a system is the interaction of that systemwith a bath.is connection will be used several times in the following,in particular in quantumtheory.When a record is produced by a machine, the 'observation 'is usually called a (generalized)measurement .Are you able to specify the bath in the case of a person looking at a landscape?Challenge 1145 n"