Seeing Like a Finite State Machine

There are two notable problems with machine learning. One ... it is not nearly as universally effective as the mythology suggests. The other is that it can serve as a magnifier for already existing biases in the data.... When this data is then used to make decisions that may plausibly reinforce those processes... the bias may feed upon itself.

crookedtimber.org/2019/11/25/s

@dredmorbius Seems like this same problem potentially applies to a great many uses of technology to amplify authority. It also points to a strategy against it: jamming, adversarial ML, deliberate false positives, etc.

@freakazoid The tendency of technology to serve as a power multiplier *is* something that concerns me.

I'm not entirely convinced it's a universal tendency, and tech can absolutely disrupt *existing* powe centres.

It just seems to usually replace them with one even more powerful.

A possible counterexample is the devolution of monarchic power in England, starting with the Magna Carta, onwards. Generally to a wider class of nobility, but since ~19thc, also commoners.

@freakazoid Which raises the question of what technologies act to *devolve* or *distribute* power.

I'm not sure I have any answers there, but it's an interesting question.

I'm also fairly certain that the Internet probably *isn't* one of those, or at least, isn't unambiguously a power-devolving technology.

@dredmorbius Maybe it would be useful to ask what properties such a technology should have?

Solar and wind: usable by individuals, independent of the grid.

PCs: usable by individuals independent of any other computer.

Mobile phones: locked-down accessories to a network that can only be operated by a huge company.

@freakazoid A notion I've had is that monopolies and power virtually always:

1. Are based on network structures -- physical, virtual, or a mix of both...

2. With either central, powerful, or privileged nodes...

3. That are not held in common or operated for social benefit and under common review...

4. Often with asymmetries of immunity or impunity of action. Think 'stand-off capability' in a military sense.

Some of these may require some mental gymnastics to visualise, but seem to exist.

@freakazoid Power (in its political, social, and economic senses) is surprisingly poorly understood and discussed, though I'm a latecomer to much political theory and may have missed much.

John Kenneth Galbraith (Anatomy of Power) names three forms: compensatory, condign, and conditioned (or money, force, and ideology/charisma).

Gavin Kennedy has a balance-of-power / forces diagram.

There's the notion of BATNA: best alternative to negotiated agreement, in contracts.

@freakazoid The essence of power is being able to apply a small effort to achieve a large outcome.

The Precinct Scene from Malcolm X is a ... powerful ;-) illustration of that:

invidio.us/watch?v=iwGojrTKWvI

Here, Malcolm's power comes from is personality, calm, the support of others, and a high level of discipline among that group. When it is time to dissipate, a simple turn of his wrist is all it takes.

(How much that was embellished for cinema, I don't know, but it's effective storytelling.)

@freakazoid What's interesting in Western traditions is that a key metaphor for power refers to horses: the reigns of power. (Effectively, the mechanism of communication to a more capable beast.)

China has been described as evolving from a hydraulic civil works project (its history of dams and canals dates back thousands of years). I'm curious if there is any notion of gates, valves, bridges, etc., within Chinese concepts of power (I've not found any).

@freakazoid @freakazoid I've come to realise that the concept of a gate, or switch, or vacuum tube ("valve" in British English), and transistor, are precisely this mechanism of applying a small change to produce a large effect.

There's Cooper's Law, "All machines are amplifiers". Not strictly true, but illuminating.

There's also the notion of _transforming_ power or motion, through simple machines (ramp, lever, screw, pulley, gear), or through converters (electric motors, lights, speakers...).

@freakazoid Sometimes power comes from the capacity to act or control with exacting precision, not merely brute force.

(In engineering: the whole science of precision and control, see Simon Winchester's recent book.)

The whole "power is networks" notion may simply reflect the fact that human structures themselves are networks -- hierarchies, webs, trees, chains, stars, and meshes -- and the power structures that _correspond_ to them match this. Though I think it's more than that.

@freakazoid I've also been rooting around in various ontologies -- industrial classifications, encyclopaedia organisation, library cataloguing schemas, and the like. And whilst there's some hazard of path-dependency following conventions and ontological artefacts, I find that the things I tend to think of as "networks" often get lumped together. Not necessarily universally, but at least within a given sub-class.

Dewey Decimal 380s classification, e.g.

@dredmorbius The definition of a machine is a thing that amplifies human work, so it is definitely an amplifier. A network is a thing that connects human will to other humans and/or to machines, so it makes perfect sense that power is in the form of networks.

This is also why writing and then printing were so important. Word of mouth was limited in time and distance and required human effort each time it was repeated, while the written word can be read widely, repeatedly, and for a long time.

@freakazoid On writing / communications, you might enjoy:

*Records* involve variations in _space_ created by a _writer_ on a _medium_ which are transmitted through _time_ to a _reader_.

*Transmissions* involve variations in _time_ created by a _transmitter_ in a _channel_ which are transmitted through _space_ to a _receiver_.

There's a symmetry here. I've not seen it noted other than by me (though suspect it has been).

@dredmorbius The physics nerd in me won't let me accept these as different things. Space and time aren't separate things. There is always a channel. There can be an arbitrary number of receivers at arbitrary points in the future. Sending signals through long lines used to be used for storage (see: delay line memory).

@freakazoid The space-time continuum is an interesting element, as is the mass-energy one.

Though another part of me says "a false model which proves useful is useful". The distinction between signal/record and time/space seems significant, relevant, and useful.

And bringing up mass/energy: records tend to be encoded in _matter_ whilst transmissions propogate through _energy_.

The (interesting) gray zone where these distinctions blur does deserve attention IMO.

@dredmorbius Matter and energy aren't different things, either. Since c is just an arbitrary constant that arises from our use of different units for space and time, you can just use the same unit for both, set c to 1, and the equation becomes E = m.

But I think what matters to records vs transmissions is *intent*. A transmission is intended to be ephemeral and received within a narrow window of time.

@freakazoid While c has an arbitrary _magnitude_ that's based on our choice of measurements, it has a _nonarbitrary_ set of *units*.

That is, 'c' is the space-time mapping between matter and energy.

@dredmorbius I've expounded on these things at length, but I think you've missed the threads. You are exactly correct that c "maps" space to time and matter to energy. But "units" would be things like "meters" and "seconds" which are most certainly arbitrary. The thing that's not arbitrary is called "dimension", which are generally labeled L for length and T for time.

But according to relativity, the axis we label "T" and axes we give the dimension "L" are dependent on the observer's velocity!

Follow

@freakazoid OK, this I've got to think about, though you make some good points.

I'm wondering to what extent "units" and "dimensions" are equivalent or at least exchangeable.

On space and time as being _only_ dimensions ... I have issues.

Length itself can be conceived of as three dimensions. Or as polar / spherical coordinates, with degrees and azimuth.

Travel in space is not directionally limited. Travel in time _is_. Alternatively: space is signed, time is unsigned.

@freakazoid The fact that the _amplitude_ of length is dependent on velocity doesn't change the fact that the _dimension_ itself is not.

Again, c remains as a units, _or if you prefer_, a _dimensional_, equivalence between mass and energy, involving time and space.

Also: the _units themselves_ of c are *independent* of velocity. c *is* c for *all* observers. 😺

Pick whatever they are: m/s, furlongs/fortnight, smoots/scruple, Whatever the units, the amplitude remains a constant.

@dredmorbius Not true! This is why the ability to move faster than light would allow time travel, and why simultaneity is not a concept that can exist in relativity. Acceleration is a *rotation* in space-time.

@freakazoid *IF* you can move faster than life *THEN* you can travel (backward) through time.

That's ... a substantial "if"....

I think I kinda get acceleration as rotation, but do you have a guide / visualisation?

@dredmorbius Actually being able to move faster than light is not necessary to my point, though; the fact that moving faster than light would enable travel through time is just a consequence of the geometry.

There is a good visualization of both the rotation ("scissors-like") and the nonexistence of simultaneity at the bottom of phys.libretexts.org/Bookshelve .

@dredmorbius The thing I will say is that the speed of light does make for an objective barrier between "space" and "time". And the ratio between the two will remain the same. But the distance or time an observer measures depends on their velocity relative to the thing they're measuring. It also depends on the curvature of space-time due to gravity. Which is why if you were falling into a black hole you'd see the event horizon receding from you while an outsider would never see you cross it.

@dredmorbius But now that I think about it I think I get what you mean by constant. Yes, the magnitude is a thing that's constant in a given reference frame; it's the numerical value itself I meant was arbitrary. I was only talking about your use of the term "units".

@dredmorbius In SI all length units are defined in terms of the distance light travels in a given amount of time these days. The base SI unit of time is the second, which is defined as a large multiple of the inverse of the frequency of a cesium oscillation.

There are natural units, too, obtained by setting various constants to 1, usually Newton's constant, Dirac's constant, and c. The natural unit of time is the Planck time, T_p or just T. The other units (L_p, etc) are derived from that.

@dredmorbius Nobody's been able to find any particular significance to these units yet, though, other than that the Planck time is shorter than any time interval known, the Planck energy is higher than the mass of any known particle, etc. So they could represent the "resolution" of the universe, but I'm aware of no evidence for this yet.

@freakazoid And the Planck mass is surprisingly macroscopic -- about 0.02 grams. About the mass of a flea egg.

en.wikipedia.org/wiki/Planck_m

@dredmorbius Right, because it's as much huger than the mass of the subatomic particles as the Planck length is smaller than their size. It's the only one that is close to being a useful magnitude for human use. Well, except the "Planck charge" I guess, which is the inverse of the square root of the fine structure constant IIRC.

Sign in to participate in the conversation
mastodon.cloud

Generalistic and moderated instance.