Eye photoreceptor

Photon induces a cascade [spreading activation]

https://en.wikipedia.org/wiki/Photoreceptor_cell
[ Like distributed consensus ] [Or like metastable system]

This is an amplifier [later there is averaging]

[ amplifiers can violate 2nd law ]

Amplification is needed to go digital (e.g. for nerve impulses) [i.e. the minimal version of A-to-D converter[

Piston / pressure measurement

This is an averager

Clocks

Where do you land relative to a counter?
[ cf interferometers etc. ]
AKA how many events do you have, compared to a standardized system having events?
[ counting events per unit time ]

Base Units

[ may be old ]

https://en.wikipedia.org/wiki/Measuring_instrument

Are Measurements Ultimately Events Per Unit Time?

First you prune to decide which events you care about ; or perhaps you aggregate events...
What macroevents happened per unit time
[ Claim: macroevents can have extent in time, space .... and branchial space ]

Macroevents are the results of coarse graining

Observers prune information, but then preserve it for a while

Reduce number of dof; increase time persistence

Basic Structure of Measurements

Collecting a limited number of dof, and putting them in a form where they can be remembered.
A nonrelativistic eye collects photons from a certain region, amplifies them and records them. Inside the measuring device, there’s rapid interaction compared to the dof outside.
Relativistic eye: the dof inside the eye are knitted together by lots of light signals; the ones from outside come in through a much more limited number of light signals.
cf piston causal graph
Minimal model of the observer is a single coherent timelike event stream. Although to “form a definite memory” it’s got to lots of events.
Consider some event chain as “the observer”, and look at other events relative to that.
For a piston in a gas figuring out pressure, the only relevant information is events per unit time on the wall... Has to compare those events to a clock.
When a measurement happens, the observer has to be largely unchanged ; changing only by having something that records a persistent memory.
Pick the chain of events that is supposed to be the observer; that’s your world line, orthogonal to the reference frame.
So now we can attribute a time t to the event as (t+ + t-)/2 , and the spatial position is (t+ - t-)/2.
With a single worldline we only get to know 1D distance to an event. Just like a telescope imagining in 2D only knows the direction to stars, not their distance. We need to use parallax to find the distance.... [cf triangulating a supernova requires multiple detectors (e.g gravitational wave detectors, etc.)]
<< Is there something about our sensory apparatus that makes us perceive 3 dimensions? >>
< cf. how do we unravel a spacefilling curve>
Our perception of 3D is somewhat exploded by the internet.... [because who knows where the servers are...] [if we lived inside a computer, why would know about 3D?] [It would depend only on the connectivity of the network]

General principle: some set of events can be elided together to make “an observer” as an integrated thing

E.g. stat mech: the observer is aggregating some number of dof’s and eliding them to a single “result of observation”
Look at MWCG ; now an observer is a bundle of timelike curves in the MWCG
Every time an “observer bundle” is formed, it has a common starting event; i.e. all parts of the observers are causally connected to some origin
[The parts of the bundle interact a lot, which is why the whole bundle can be considered “rigid”, and elided (like a rigid body can be described just by center coordinates and orientation) ]
The lowest-level version of the MWCG involves no merges; any merge must be “in the eye of the observer”
Consider threads in MWCG as computer threads; forking when there is a branch in the MWCG
(it could be that a local region of causal graph is identical to another local region)
When can you use “shared common expressions” to prune the threads?
Pure tree ;;; then add equivalences
https://www.wolframinstitute.org/bulletins/2021/02/multiway-turing-machines/

Observer is a bundle of worldlines in which certain outside events are amplified to affect lots of these worldlines (i.e. many worldlines depend on the outside event; the effect is the persistent memory) but there is still persistence in the bundle; the observer can then be “equivalenced down” to say that a single outcome occurred

Multiple scales:
scale of whole system
scale of the whole observer [carrier of the memory]
scale of an individual memory
Scales in terms of time; scales in terms of dof

In stat mech, worldlines are at different places in space, but have the same state of motion

Many elementary proto-observers are being combined...

In relativistic case, the different places in space are close together

In quantum case ... there are pre-proto-observers

Bundled branches are a “multi-observer”
Multiple branches running the same program... We can readily conflate this to a single branch
If an observer is approximated by a single multiway worldline ... then the observer can perfectly well think that they are not branching....
When the observer is a bundle of worldlines ... it becomes implausible for the observer to think they branched (?)
As soon as you coarse-grain in branchial space, you have irreversibility....

Rulial Observers

You can fish out of the ruliad any apparent laws just by making the appropriate interpreter for a universal computer

What is the origin of the “stability of experience”?

It just takes a lot of events (i.e. a lot of time) to “get to somewhere different”

Claim: rate of change of the observer is slow compared to the rate of change of the universe

[Imagine we map our whole view of the universe onto a Turing machine .... it would ultimately work, but not in real time]