Eliezer Yudkowsky on Politics

(This article was reprinted in the online magazine of the Institute for Ethics & Emerging Technologies, December 20, 2016.)

In our previous post, we examined how prospect theory helps explain why so many American voters were willing to risk voting for such a manifestly unqualified candidate for President as Donald Trump. Of course what citizens who are willing to take these risks fail to understand, as the artificial intelligence and decision theory expert Eliezer Yudkowsky writes on his Facebook page, is “how there’s a level of politics that’s theater and a level of politics that’s deadly serious.” For example, it’s deadly serious when a President talks about scrapping the NATO alliance or using nuclear weapons. In such cases, you would hope that competent and conscientious people exercise power in the international relations realm.

Unfortunately, some people don’t seem to understand the difference between entertainment and reality. Yudkowsky offers the example of those who believe the moon landing was faked. From an educated perspective, this belief is self-evident nonsense, and I understand that TV shows that promulgate such nonsense are entertainment. But some people don’t know this; they don’t know there’s real science that allowed us to go to the moon, a serious level underneath the entertainment level. (There the entertainment level of having some ignorant evolution denier debating someone on TV, and the serious level where biological evolution underlies all of modern biology.) But often the scientifically illiterate only know the world that comes to them from their entertainment bubble.

Similarly, many people can’t differentiate political theater, Level A, with the deadly serious part, Level B. Perhaps they think that Level A is all that exists and there is no deadly serious politics, no Level B. They are mistaken.

But the Level B in Washington DC, the issues that people take seriously unlike insider trading, is also not just sociopaths reacting to disasters that are *so* bad that their own personal hometown might get a nuclear missile. The Level B does contain more stuff than that … it *is* the level where you worry about things like the stability of the Europe-Russia border, not because a journalist is going to clutch their pearls in offense because you don’t seem concerned enough, but because you actually care about the stability of the Europe-Russia border. Yes, there are people in Washington DC like that.

So there is a deadly serious level of politics that demands equanimity. In this context, Yudkowsky notes: “Perhaps there are dozens of other cases where a country elected an impulsive, chaotic, populist leader and nothing whatsoever went wrong.” But in such situations something could go terribly wrong. Yudkowsky recalls playing a National Security Decision-Making Game with about 80 participants and finding out how much strategizing it takes to avoid oblivion.

By the end of NSDM, I left with a suddenly increased respect for any administration that gets to the end of 4 years without nuclear weapons being used … I left with a greatly increased appreciation of the real skill and competence possessed by the high-level bureaucrats … who keep everything from toppling over …

I think that a lot of the real function of government is to keep things from toppling over like they did in our NSDM session and that this depends on the functionaries including the President staying inside certain bounds of behavior––people who understand how the game is supposed to be played. It’s not always a good game and you may be tempted to call for blowing it up rather than letting it continue as usual. Avoid this temptation. Randomly blowing it up will NOT end well. It CAN be so, so much worse than it already is.

The system isn’t as stable as it might look when you’re just strolling along your non-melted streets year after year, without any missiles ever falling on your own hometown. I don’t even know how much work it really takes to prevent everything from falling over.

Pursuant to the above, Yudkowsky argues how dangerous it is to have a President like Trump. It is bad to be ambiguous about who will defend who for example. Both world wars in the 20th century began because of such ambiguities. And it is deadly dangerous to wonder why we shouldn’t use nuclear weapons. This leads Yudkowsky to muse about the relative importance of otherwise important issues:

Like it or not, there is in Washington DC a perceived difference between ‘committed sexual assault’ and ‘violated the system guardrails that prevent World War III’ … Maybe you wish that Washington DC culture would take sexual assault more seriously, as something deadly serious in its own right … instead of some people laughing it off, some people being frankly offended, and everyone in Washington DC tacitly understanding that this is not one of the issues that everyone has agreed to take deadly seriously even when no journalists are looking.

Maybe you look at that and conclude that this ‘deadly serious level of politics’ thingy does not respect your own values and priorities. Maybe you conclude that the kind of political issues people are fighting over theatrically in the newspapers are, yes, every bit as vital to you as that so-called ‘deadly serious’ stuff even if a lot of other people are treating them as entertainment.

I think you’re making a dreadful mistake. Scope is real. If you ever have to choose between voting a convicted serial abuser of children into the Presidential office—but this person otherwise seems stable and collected—versus a Presidential candidate who seems easy to provoke and who has ‘bad days’ and doesn’t listen to advisors and once said “Why do we have all these nukes if we can’t use them?“, it is deadly important that you vote for the pedophile. It isn’t physically possible to abuse enough children per day over 4 years to do as much damage as you can do with one wrong move in the National Security Decision-Making Game.

An evil but the same NSDM player is far, far less dangerous than an impulsive one who doesn’t care all that much about what the rules of NSDM are supposed to be.

I suppose you can now see why Trump worries a deep thinker like Yudkowsky. As for me, I basically agree with everything Yudkowsky says here. We should remember that survival is a prerequisite for the existence of any beauty, joy, truth, or goodness in our fragile existence. Civilization bestows so many benefits compared to the state of nature most of us have long since forgotten. Civilization itself is fragile; it can end at any moment. The politics of Level B is deadly serious and our very survival depends on serious, knowledgeable people occupying positions of extreme power.

Part 2 

(This article was reprinted in the online magazine of the Institute for Ethics & Emerging Technologies, December 22, 2016.)

Our two previous posts showed how prospect theory in behavioral economics explains why so many gambled on Trump, and why the artificial intelligence and decision theory expert Eliezer Yudkowsky thinks that this was a mistake. In a post written the day before the election, Yudkowsky expanded on both themes, providing a simple explanation of  how many of the gamblers reasoned:

Life (in Alabama, let’s say) used to be good. Then it got worse. So something is going wrong. Something must be making life in Alabama worse than it used to be a couple of decades earlier. Some malevolent force is pushing life in Alabama away from its natural default state of goodness.

Then it would be wise to do something, anything, differently. Like whacking your malfunctioning microwave with your hand, in hopes that you shake loose whatever component is in a rare state of malfunctioning, and the microwave goes back to its default state of working correctly.

On this perspective, most possibilities … are pretty good … So if life … seems bad, there must be some unusual factor that’s forcing things to go poorly … In which case there’s a lot to be said for overturning the table and doing *anything* except more of what you’re currently doing … The most important thing you need in a President is that they not be part of the same malevolent structure that has repeatedly punched you in the nose.

Then there’s the other perspective:

Most countries in the world aren’t as nice as the United States is right now. Venezuela used to be an up-and-coming country with one of the fastest-growing economies in South America. And then they elected an impulsive populist leader who made a few decisions he probably didn’t think were that bad at the time, and now Venezuela is on the verge of being a failed state.

The good things are fragile. It takes hard work to preserve them, and even people who try to do that sometimes fail … History shows the kind of global prominence the US currently has, can fade very very quickly if a country makes a few wrong moves.

Even “relative” prosperity can fade very very quickly … Most ways of reaching into your microwave that’s been seeming a bit elderly, and randomly switching up the circuits a bit, will very very rapidly cause your microwave to work worse. A lot worse.

The point is that the modern world is a very dangerous place. Forget to tell allies you’ll defend them, and wars start. Mistakenly tell Saddam Hussein you don’t care about Kuwait, and he will believe you and invade there. As Yudkowsky puts it: “Being President means standing on a shaky platform over a pit of radioactive lava, on which platform also happens to rest the United States and often the rest of the world.”

If your choice is between two qualified candidates then, by all means, choose the one you think is better for you—assuming both candidates understand serious politics! But if one of them doesn’t, then it is madness to choose the madman. In other words, your current situation in the USA may not be great, but it is probably better than the catastrophically bad outcomes that might result from putting people in important positions who don’t understand the seriousness of politics.

Pumping up the entropy doesn’t shatter a fragile malevolent thingy and let us go back to the normal good days. It obliterates the careful moves that barely manage to achieve the meh results you see around you, and dumps us into the boiling lava underneath. That’s what happened to Venezuela …

In short, things can quickly get much worse.

And that’s really and actually true, despite your sense that things like that aren’t *allowed* to happen, they can’t *really* happen, not *here*. Not to *you*. Not in *real life* as opposed to the 1930s. That’s all only in history class, which is part of the same fictional continuum that includes movies and television.

You can read endless phrasings of those very words, from people in various countries that went downhill, about how they thought it couldn’t happen to them. “That couldn’t *really* happen here, could it?” is a common refrain, historically speaking, from just before very bad things happen to countries …

Of course, some smart people still don’t think Trump and the Republicans could do that much damage, and maybe it will turn out ok. But don’t expect that existing political structures will restrain him. Sure Obama had limited power, but he played by the rules.

The lesson of history is that populist strongmen … can happen here, they can happen anywhere, it happens all the time, not to aliens but to populations of human beings pretty much exactly like the population of human beings surrounding you …

… Lesson one of history: Populist strongmen *fire* the senior bureaucrats who don’t obey them and replace them with loyalists. And that works. The strongman does successfully consolidate power and he is obeyed henceforth, even by people who theoretically shouldn’t obey him, even when it is theoretically against the law.

Lesson two, you’d be *amazed* at how fast senior bureaucrats capitulate to populist strongmen. I was amazed at how fast the existing Republican party structure rolled over for someone who’d cheerfully slit their throats, back when Trump was running for President with, God help them, the support of the Republican leadership … the history books repeat over and over … the story of the surprisingly fast capitulation of government bureaucrats and key social figures to the strongman …

Still, many people, especially white males, don’t think it can get that bad. (Non-whites and non-males better understand how bad it can get.) Surely somebody will stop it. Somebody will stop Trump from getting the nomination. Some military officers will stop the use of nuclear weapons. But even if they did, Trump could force another to carry out his orders. Here are the lessons Yudkowsky takes from Trump not being stopped thus far:

1) Goodness is fragile; 2) What you have can be taken away. Surprisingly quickly, if not quite overnight; 3)The nebulous people and forces that are supposed to stop bad things from happening, won’t. It *can* happen here.

And, perhaps most importantly, Yudkowsky compares the situation to what he knows best, artificial intelligence:

AIs with random utility functions will not turn out pretty much all right. Even AIs that somebody tried to make go right, but screwed up in any of a dozen ways, will not turn out pretty much all right …

Your sense that this current world is true and enduring and steady, that it has existed forever in the past and will always exist in the future, that modern humans have existed forever and will always exist and that developments like Artificial General Intelligence are part of the continuum of movies and fairytales, is only slightly less blind than thinking that the 1940s were too far in the past to be real.

There’s no nebulous group of competent, well-intended AI scientists who will make sure that everything goes reassuringly well … things are allowed to not turn out all-right. … It can enter into your immediate reality instead of being safely on television or in history books or somebody else’s Facebook wall.

Yudkowsky concludes with some of the most chilling words I’ve ever read:

You could wake up on November 9th, 2016 to find that the United States as you knew it has ended …

It’s *allowed*. Learn from that now, before Wednesday, while you can still ponder that awful uncertainty. No nebulous surely-someone prevented Donald Trump from gaining the Republican nomination, no nebulous surely-someone stopped half the country from voting for him, and if November 8th goes poorly, no nebulous surely-someone will prevent all the other things that Donald Trump goes on to do.

And none of that will matter, on some later day when you don’t wake up at all. Because there were no nebulous forces that swooped in and saved the day. Because the reassuringly powerful and competent and benevolent people who are supposed to make sure that things aren’t allowed to get that bad, do not actually exist. This world of ours isn’t so strong, just like the United States proved weaker than you expected. What you see is all there is—maybe a little less.


(I don’t know Mr. Yudkowsky but I taught his “Creating Friendly AI” to computer science students at the University of Texas at Austin. I thank him sincerely for his various intellectual contributions.)

Liked it? Take a second to support Dr John Messerly on Patreon!
Become a patron at Patreon!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.