But I’ve started watching Apple TV’s “Foundation” series based off of the classic Isaac Asimov books where mathematician Hari Seldon develops his “psychohistory” (a sort of mishmash of chaos theory, predictive analytics, and sociology) that predicts the collapse of the Galactic Empire.
A similar theme was covered in HBO’s Westworld where they had various supercomputers predicting the end of civilization in absolute terms as well as the film Downsizing where scientists predicted climate change and resource depletion would collapse civilization in a few hundred years.
So that sort of got me thinking about our current modern use of AI and predictive analytics to try to predict trends or catastrophes such as climate change, economic forecasting, and so on.
I suppose the “Great Debate” is do people believe that these sort of mathematical models are feasible, practical, or even actionable?
In fiction, they are always sort of presented as a done deal with uncanny accuracy. Even if (in the case of Foundation), the math behind it is largely unproven and only understandable by a handful of people.
In practical real-world terms, I suppose the question is to what extent we upheave civilization now to head up some potentially more massive upheaval 10, 50, 100 years from now and to what extent do we trust math that only a few experts really understand.
Having studied a bit of math, I can see the usefulness of creating models to try to determine best outcomes. But it’s sort of like predicting the weather a year from now. I know I should have an umbrella in my household. That doesn’t help me know if I need one today.
This is why I found the Foundation series so frustrating. It’s fabulous storytelling, but the preposterous idea that a model could predict the future so precisely just constantly took me out of it. It’s easy so suspend disbelief about something like FTL for the purposes of a story, but I just found it impossible with this.
The short stories were ok. As they assembled into the trilogy and then the series I found difficulty accepting it as well defined universe. The existence of an algorithm that can predict human events like that means there was underlying programming of humans. It’s been so long I don’t recall many details but don’t recall that really being examined.
I found that a bit frustrating as well:
“Our mathematical model predicts the end of Galactic civilization in around a few hundred years, followed by between one and ten thousand years of dark ages.”
“That sucks. Can we stop it?”
“No”
“Are we sure the model is correct?”
“Maybe. It’s a new model and just the one guy who built it kind of understands it. Also, it’s not complete.”
“So is there anything we should be doing (or not doing) now?”
"Can’t really say. The model doesn’t make predictions at that level of detail.
What we can tell you that over the next several years and decades, you will experience
a massive terrorist attack
a religious uprising
some sort of ideological shift
some other massive catastrophe
and so on and so forth."
“Ok, great. Normal ‘running a galactic empire of quadrillions of people on a hundred thousand planets’ shit. Thanks for the heads up ‘Nostradamus’.”
In this particular case, it seems like the Emperor made the right decision in “exiling” Hari and his doomsday cult to some backwater to go work on their math and dig their bunkers until they came up with something more useful.
Speaking of which, imagine for a moment that you’re a college student in 1901 — Harry Truman, say — and you get asked by the Wright Brothers to opine about heavier-than-air flight; also, can the atom be split? If so, will airplanes be used to drop atomic bombs on cities in World War II? Also, will you get the chance to watch on TV as a guy plays golf on the moon?
Your point is that FTL, like this or that other innovation, makes certain stories possible; I guess I’m asking: how can someone predict (centuries in advance) which stories will play out for a society if they can’t predict (decades in advance) the science that’d make those stories possible?
What could Truman — or anyone else — say about plutonium decades before it gets discovered?
Foundation is more of an exception in that it predicts comparatively long term trends over an extended period. Most other fictional variants are predicting a single, near future crash or event, so they’re (probably) a much more probable example of a possible “real” algorithm. But, I suspect the issue with a real version is our ever popular GIGO (garbage in, garbage out) where the model is reasonable, but only workable based on the assumptions of the humans programming the model.
Foundation (the novels) was also built on the assumption that yeah, individuals couldn’t be mapped, but would only contribute “noise” as the actual trends would remain unchanged. But as already pointed out by @xtenkfarpl a sufficiently abnormal (or exceptional) individual could toss those assumptions out. And (spoilers!)
Foundation Novel Spoilers
It was bullshit from the start as the later novels showed. It took an entire planet of Hari’s students, working behind the scenes to “force” the chosen time-line to remain in complaince.
No—and I say this as a white man—I just think it’s a white man power fantasy about not only controlling the future in one’s own lifetime, but guiding the fates of untold trillions of humans into the distant future. It’s got all the viability of “separate but equal” or “a benevolent dictatorship” as a means of maintaining social harmony.
I believe the idea was that you couldn’t predict the actions of individual people or even small groups but that above a certain level the actions of individuals averaged out and you could predict the actions of the masses.
I believe Asimov’s real plot was to retell the collapse of the Roman Empire and the rise of western civilization in a science fiction setting. He invented pschohistory to strengthen the plot by adding the idea that some people could see these events as they were occurring.
You can’t predict the future, no matter how smart you are, because the real world is fundamentally unpredictable. There are no algorithmns that can predict the unpredictable.
The thing that makes the world unpredictable is complexity and quantum dynamics. Comolexity means feedback, non-linear responses, and as a result extreme sensitivity to initial conditions. Sensitive enough that even completely random quantum fluctuations matter.
Look at the major changes we have gone through in the past two decades alone. Basically a series of unpredictable Black Swan events have hit us and apparently fundamentally changed us in some ways. There are billions of unseen things that happened that also changed us, and are always changing us. And a decade from now we’ll look back at our predictions today and laugh.
No one can predict the future, except in some broad strokes by following trendlines and hoping they don’t become discontinuous. That goes for macroeconomists, the CBO, mystics and fortune tellers. None of them do better than chance in the long term. Looking back at economist’s predictions for the future is always good for a laugh.
I couldn’t get through the Foundation books for that reason. The core premise was just so wrong I couldn’t get past it.
Models running on supercomputers, computational methods, “psychohistory” if you want, in the field of social sciences is a real thing; there has been a push for it for many years already and that has not ended. People, and funding agencies, believe these sort of mathematical models are feasible and practical. But that does not equate to knowing the future “as a done deal with uncanny accuracy”. Still, would it not be nice to be able to mitigate the effects of recession, famine, epidemics?
What is wrong with predicting the weather, as it were?
You don’t need AI. Calculating the expected end date of humankind is a trivial application of probability theory and Bayesian reasoning.
Humanity gained the ability to destroy itself, roughly speaking, in 1945 with the advent of nuclear weapons. So far almost 80 years have gone by without an extinction event. So we’re 0 for 80. Using the rule of three, we can be 95% sure that the per year odds of extinction is somewhere between zero and 3/80 (0-3.75%). So at the lower bond we’ll never go extinct and at the higher bound we have a 50-50 chance of making it past July 2043.
The unauthorized sequel to Foundation - “Psychohistorical Crisis” (by Donald Kingsbury) has a more realistic look at psychohistory, by looking at the question of what happens when more than one group has the ability to model what’s going to happen (Michael Flynn’s “Country of the Blind” does the same thing for the present-day).
Asimov himself did a jokey short story about a guy who gets rich upon programming a computer to predict whether stocks will go up or down; it works “because things that now take place are the same as have taken place in the past, so that the response is the same.”
And, in that context, he figures there’s nothing new under the sun: “Go through history and you’ll find that there are only changes in detail.”
Eventually he’s the richest man in the history of the world, with plenty of political clout and a computer that keeps poring over more and more information to give him more and more international oomph — until the computer finds itself struggling to process a completely new factor in human history: nothing like it has ever existed, and so weighing its impact as a variable is an impossible task that leads to destruction and chaos.
Because, of course, every other historical event it tries to extrapolate from had one thing in common: nobody was ever influencing world affairs with the aid of a prediction-making computer.
