Future Tables

We don't want future visions, we want future tables

This research note is part of the Clockless Clock book project.

Sidebar: This note had been sitting in my drafts for months, as an unfinished line of thought, but I just read an interesting essay by Josh Stark, Atoms, Institutions, and Blockchains, (recommended as a paired read), which makes very similar points using a different frame. So I thought I’d flush this out despite the unfinished state, in the interests of provoking a broader conversation on the general theme.

A table is a flat physical structure to support work material in a convenient position. A table in a document holds elements in fixed spatial relations to each other, using a grid of rows and columns reminiscent of the game boards they are likely descended from. You can "look up" a table because its elements won't change in weird ways between lookups. The etymology is somewhat ambiguous but revealing:

Both the French and Germanic words are from Latin tabula "a board, plank; writing table; list, schedule; picture, painted panel," originally "small flat slab or piece" usually for inscriptions or for games (source also of Spanish tabla, Italian tavola), of uncertain origin, related to Umbrian tafle "on the board."

The sense of "piece of furniture with the flat top and legs" first recorded c. 1300...The meaning "arrangement of numbers or other figures on a tabular surface for convenience" is recorded from late 14c. (as in table of contents, mid-15c.).

We use the verb form to refer to the deferral of decisions. To table a decision is to reschedule it in a more distant location on a calendar, itself represented as a symbol table in two-dimensional textual space. The future of a procedurally stabilized future is a set of scheduled or rescheduled events and decisions on a time table.

Physical and informational tables are clearly related. Both present elements of a functional environment in stable positions relative to each other, and to the surrounding environment. Things on a table stay where they are unless deliberately repositioned. A table enforces respect for what in AI is called the closed-world hypothesis: a world where things don't change unless acted upon by an agent. It mitigates the frame problem of open worlds through design.

In fact, a physical table is the prototype of all closed worlds. It is an entity that frames, bounds, and stabilizes an open world, turning it into a closed world with some pseudo-deterministic properties, where I define pseudo-determinism as the logical dual of pseudo-randomness: a property of events that we can usefully treat as functionally equivalent to determinism in many applications.

Chess and Go are pseudo-deterministic games played on tables (on tables) where nothing changes unless a player makes a move. It is interesting that in games of indeterminacy, such as card or dice games, the indeterminate elements acquire pseudo-determinacy only when they enter game-play on the table (dice being rolled or cards being placed on the table). Table-top games are somewhat less pseudo-deterministic than board games in part because they accommodate two levels of tabling. Video games can accommodate many layers of virtual tabling (levels and such).

Note that any dice-like indeterminacy accommodated by tabling is known-unknown, modeled indeterminacy, injected in a highly controlled way. A situation where an upset player sweeps the game material off the table is not modeled (hence the pseudo-), but is not zero probability either. It’s a sort of local toy-black-swan element in the phenomenology. The probability (and/or disutility) of such a toy black swan event is simply low enough to ignore; it is below our indifference threshold.

Tables in space create determinacy in time. Even a dinner table, with place settings around the perimeter, with silverware arranged in the order of courses, is an engine of pseudo-determinacy shaping events for an hour or two. Styles of dining with less pseudo-determinacy, such as family style, introduce a frisson of excitement to the extent you're not sure what you're going to get, when.

The function of any table, then, is to create pseudo-determinacy, of events in time, which leads to orderly expectations.

The latent potential of things on a table -- be it a dinner table, a look-up table, or a time-table -- translates to orderliness in the closed world of events it generates and circumscribes in time. The presence of a table indicates a certain low-entropy, high-predictability trajectory for all events within a certain protected scope. To the extent the energy required to navigate in-scope events is the energy required to anticipate and defend against out-of-scope events, a table lowers the energy demanded by the future it enables. In Deleuzean terms, a table turns a smooth future into a striated one.

A bus timetable is a simple example. If you plan on taking a bus, and it is on time, you don't need to expend energy worrying about routes, traffic jams, the availability of cabs or rideshares, and so on. If you are running late, you might have to run -- a high-energy burst -- to make your bus, but on average, a bus timetable indicates the presence and availability of a highly determinate, low-energy, low-entropy future for you to participate in. This future is not just predictable and low-energy for the events within it, it requires energy to disrupt. To disrupt a single arrival on bus schedule requires something like a traffic snarl. To disrupt many requires something like a mechanical breakdown. To disrupt the bus timetable of an entire city for a day requires something like extreme weather. To disrupt it for weeks requires something like a war.

So a timetabled future is something like a time-crystal of highly predictable events running within a stable, low-potential groove. Adjacent-possible futures flow into it. If you and your neighbor, who both commute by the same hourly bus to the same office, leave home at 7:50 and 7:55, and get on the same 8:03 bus, your futures have converged. You'll probably ride the elevator together when you arrive.

The more events there are on a time table, and the more evenly distributed they are, the more pseudo-determinate the future they represent -- assuming the potential groove (or rut) is deep enough to withstand disruptions and approximately maintain the closed-world condition. We don't randomly fly off into space because we're at the bottom of a deep gravity potential well. Events in a powerful time table don't get randomly derailed, because they're at the bottom of a deep temporal potential groove.¹

One way to understand a timetable is as a channel of notionally undiscounted events framing relatively discounted events. Though in general one must discount future events as a function of their distance in time (as in the use of exponential discounting in the forecasting of things like cash flows), timetabled frame events have a discount factor so low it might as well be zero, and they lend relative stability to linked events that are not themselves as stable.

For example, we do not bother discounting the probability of future sunrises. Sunrises are on a timetable that would take such low-likelihood, high-energy events to disrupt (such as the Sun itself going nova or aliens blowing it up), we effectively use a discount rate of 0% in thinking about them. In fact we use sunrises to count "days" and thereby construct the basic scaffolding of all time tables and discounting schedules. Other events are discounted relative to the discount factor of sunrises, notionally set to zero. Your company's projected revenue next quarter might have a discount factor of 1%, but that is relative to the implied discount factor of all future sunrises (in fact, all sunrise-grade frame events) set to 0% as a design decision.

At the other extreme, a vision is the most degenerate and fragile kind of time table, reliant on a single, powerful, non-impossible attractor event or cluster of events to inspire flows of events flowing towards itself. In Avengers: Infinity War, Doctor Strange finds just 1 future in 14,000,605 sampled futures leading to the defeat of Thanos. Since we know that that future actually unfolded five years later, assuming a horizon of five years, we get an annual discount rate of 99.9999985715%.² It's an anti-sunrise.

Unlike a time table, a "vision" like "colonize Mars" does not by itself create the low-potential timetable grooves required to draw events towards itself. Those must be created using "plans." To the extent the environment is too uncertain to plan, a higher-energy sort of improvisation must take place. To the extent necessary knowledge is unavailable, investigations must be undertaken to create it.

The next sunrise will almost certainly (which equals certainly in a pseudo-deterministic world) occur no matter what you do, and whether you live or die. Discount rate effectively zero in the real world, and exactly zero in the imagined pseudo-deterministic modeled world of the psyche.

“Colonize Mars”… not so much.

For humanity to colonize Mars by 2050, someone (possibly Elon Musk) must plan, and invent missing pieces along the way. Discount rate? If you think he has a 50-50 shot, you are using an annual discount rate of about 2.4%. Or to turn it around, you believe it is 97.6% likely the world will stay on the course to a Mars colony by 2050, year after year, thanks to Elon Musk existing. If he randomly dies, or gets too distracted with other adventures, that probability drops sharply.

Time tables generally anchor to the fates of a few anchor objects -- buses, spreadsheets, dinner courses, Elon Musks -- that themselves possess some sort of natural stability. You could say time tables are stability-contagion network effects among possible events. The fact that a bus is not going to suddenly disappear -- it has a certain reliable existential permanence that can induce and sustain subjective object permanence -- is what allows bus time tables to exist, and many other events and actions to be planned in relation to them. Within the narrow slice of future events they dominate almost absolutely, buses induce time tables. Elon Musk's future is, sadly, not quite as secure as that of the median (highly interchangeable and replaceable) bus.

The events of the world at large though, are generally not anchored to a known set of stable anchor objects that exhaustively determine, or time-table its future. So the proportion of the future structured by time tables is generally very small. The rest of the future is governed by large forces (often quiescent most of the time) that are themselves ungoverned, and cannot be easily wrangled into time tables, and are in fact capable of overwhelming what the guard potentials of the temporal potential grooves that exist for our pseudo-deterministic comfort and convenience. Over a sufficiently long period, thanks to the law of large numbers, such overwhelming of timetables by ungoverned events is guaranteed. Change, as they say, is the only true constant. The pandemic has been a good reminder of this.

This sort of thing is why the determinacy of a time table is pseudo-determinacy at best. Yet, it is clearly a distinction condition from a future unpaved by time tables altogether.

The phenomenology of tables, tabling, and time-tabling suggest a potentially powerful approach to futurism that I'll call "future tables."

A future table is an idealized time table that is large enough in scope, and with sufficiently energized guard potentials, that it can make all future events seem exactly as determinate as you might want them to be. Things you want to be certain will seem certain, things that you want to remain surprising will remain spoiler-free. An ambitious young person might want a future full of possibilities; a world-is-my-oyster future table. An old person might want a future in which their legacy is secure, and things they've come to value through life remain sacred.

In fact, to a first approximation, the sense of future time is a coherent future table. A variety of research findings³ suggest that the sense of time itself is an illusion rooted in believably firm future tables. When those unravel, the sense of time can collapse, leading to deep trauma.

The useful word anomie points to misregistration between desired and experienced certainty levels across all facets of life. We experience anomie when we find uncertainty where we want certainty, or vice versa. If that condition holds true across almost all experience, we experience a sense of no future. There is a reason why the book in which Emile Durkheim developed the concept of anomie is titled Suicide. Sufficient anomie destroys all sense of the future. It breaks future tables underlying our sanity.

The future, you might say, is mostly pseudo-determinate over individual lifetimes, and non-existent outside of the perspective of intact psyches.

All futurism seeks to address an anxiety that can only be properly addressed by a sufficiently firm future table capable of preserving psyches. A futurism is successful to the extent it is supplies good default future tables for you to customize and build on.

It need not be true however. A religion with an afterlife vision is in fact a very successful kind of future table. A caricature might be: if you do more good than bad, you'll end up in heaven. So a very robust input behavior (more good behavior than bad) lands you in an eternal future of undiscounted positive experiences for evermore. Thought experiments like Pascal's wager emerge from the powerfully distortionary effects of such afterlife boundary conditions. In such a theology, your mortal existence is like a dice-roll onto a table top where a determinate game is unfolding. You want to land in that game with the best starting roll.

For the non-believers among us, there are no perfect future tables. But we can probably do better than merely trying to wrangle a sense of the future anchored in poetic, charismatic, singular visions.

1. They can get derailed in more internally driven ways though, such as through fork events in software development. Or to go back to our board game examples, in theory, players in a dispute might replicate the game state on a new board, fork the game, and continue with other players. Forks are out of scope for this essay, but are a planned chapter in the book

2. Someone check my math plz

3. Somewhere in my notes, these do exist…. they’ll make it into the book.