Shtetl-Optimized

The blog-commenters come at me one by one, a seemingly infinite supply of them, like masked henchmen in an action movie throwing karate chops at Jackie Chan.

“Seriously Scott, do better,” says each henchman when his turn comes, ignoring all the ones before him who said the same. “If you’d have supported American-imposed regime change in Venezuela, like just installing María Machado as the president, then surely you must also support Trump’s cockamamie plan to invade Greenland! For that matter, you logically must also support Putin’s invasion of Ukraine, and China’s probable future invasion of Taiwan!”

“No,” I reply to each henchman, “you’re operating on a wildly mistaken model of me. For starters, I’ve just consistently honored the actual democratic choices of the Venezuelans, the Greenlanders, the Ukrainians, and the Taiwanese, regardless of coalitions and power. Those choices are, respectively, to be rid of Maduro, to stay part of Denmark, and to be left alone by Russia and China—in all four cases, as it happens, the choices most consistent with liberalism, common sense, and what nearly any 5-year-old would say was right and good.”

“My preference,” I continue, “is simply that the more pro-Enlightenment, pluralist, liberal-democratic side triumph, and that the more repressive, authoritarian side feel the sting of defeat—always, in every conflict, in every corner of the earth.  Sure, if authoritarians win an election fair and square, I might clench my teeth and watch them take power, for the sake of the long-term survival of the ideals those authoritarians seek to destroy. But if authoritarians lose an election and then arrogate power anyway, what’s there even to feel torn about? So, you can correctly predict my reaction to countless international events by predicting this. It’s like predicting what Tit-for-Tat will do on a given move in the Iterated Prisoners’ Dilemma.”

“Even more broadly,” I say, “my rule is simply that I’m in favor of good things, and against bad things.  I’m in favor of truth, and against falsehood. And if anyone says to me: because you supported this country when it did good thing X, you must also support it when does evil thing Y? (Either as a reductio ad absurdum, or because the person actually wants evil thing Y?) Or if they say: because you agreed with this person when she said this true thing, you must also endorse this false thing she said? I reply: good over evil and truth over lies in every instance—if need be, down to the individual subatomic particles of morality and logic.”

The henchmen snarl, “so now it’s laid bare! Now everyone can see just how naive and simplistic Aaronson’s so-called ‘political philosophy’ really is!  Do us all a favor, Scott, and stick to quantum physics! Stick to computer science! Do you not know that philosophers and political scientists have filled libraries debating these weighty matters? Are you an act-utilitarian? A Kantian? A neocon or neoliberal? An America-First interventionist? Pick some package of values, then answer to us for all the commitments that come with that package!”

I say: “No, I don’t subcontract out my soul to any package of values that I can define via any succinct rule. Instead, given any moral dilemma, I simply query my internal Morality Oracle and follow whatever it tells me to do, unless of course my weakness prevents me. Some would simply call the ‘Morality Oracle’ my conscience. But others would hold that, to whatever extent people’s consciences have given similar answers across vast gulfs of time and space and culture, it’s because they tapped into an underlying logic that humans haven’t fully explained, but that they no more invented than the rules of arithmetic. The world’s prophets and sages have tried again and again over the millennia to articulate that logic, with varying admixtures of error and self-interest and culture-dependent cruft. But just like with math and science, the clearest available statements seem to me to have gotten clearer over time.”

The Jackie Chan henchman smirks at this. “So basically, you know the right answers to moral questions because of a magical, private Morality Oracle—like, you know, the burning bush, or Mount Sinai? And yet you dare to call yourself a scientific rationalist, a foe of obscurantism and myticism? Do you have any idea how pathetic this all sounds, as an attempted moral theory?”

“But I’m not pretending to articulate a moral theory,” I reply. “I’m merely describing what I do. I mean, I can gesture toward moral theories and ideas that capture more of my conscience’s judgments than others, like liberalism, the Enlightenment, the Golden Rule, or utilitarianism. But if a rule ever appears to disagree with the verdict of my conscience—if someone says, oh, you like utilitarianism, so you must value the lives of these trillion amoebas above this one human child’s, even torture and kill the child to save the amoebas—I will always go with my conscience and damn the rule.”

“So the meaning of goodness is just ‘whatever seems good to you’?” asks the henchman, between swings of his nunchuk. “Do you not see how tautological your criterion is, how worthless?”

“It might be tautological, but I find it far from worthless!” I offer. “If nothing else, my Oracle lets me assess the morality of people, philosophies, institutions, and movements, by simply asking to what extent their words and deeds seem guided by the same Oracle, or one that’s close enough! And if I find a cluster of millions of people whose consciences agree with mine and each others’ in 95% of cases, then I can point to that cluster, and say, here. This cluster’s collective moral judgment is close to what I mean by goodness. Which is probably the best we can do with countless questions of philosophy.”

“Just like, in the famous Wittgenstein riff, we define ‘game’ not by giving an if-and-only-if, but by starting with poker, basketball, Monopoly, and other paradigm-cases and then counting things as ‘games’ to whatever extent they’re similar—so too we can define ‘morality’ by starting with a cluster of Benjamin Franklin, Frederick Douglass, MLK, Vasily Arkhipov, Alan Turing, Katalin Karikó, those who hid Jews during the Holocaust, those who sit in Chinese or Russian or Iranian or Venezuelan torture-prisons for advocating democracy, etc, and then working outward from those paradigm-cases, and whenever in doubt, by seeking reflective equilibrium between that cluster and our own consciences. At any rate, that’s what I do, and it’s what I’ll continue doing even if half the world sneers at me for it, because I don’t know a better approach.”

Applications to the AI alignment problem are left as exercises for the reader.

---

Announcement: I’m currently on my way to Seattle, to speak in the CS department at the University of Washington—a place that I love but haven’t visited, I don’t think, since 2011 (!). If you’re around, come say hi. Meanwhile, feel free to karate-chop this post all you want in the comment section, but I’ll probably be slow in replying!

I woke up yesterday morning happy and relieved that the Venezuelan people were finally free of their brutal dictator.

I ended the day angry and depressed that Trump, as it turns out, does not seek to turn over Venezuela to María Corina Machado and her inspiring democracy movement—the pro-Western, Nobel-Peace-Prize-winning, slam-dunk obvious, already electorally-confirmed choice of the Venezuelan people—but instead seeks to cut a deal with the remnants of Maduro’s regime to run Venezuela as a US-controlled petrostate.

I confess that I have trouble understanding people who don’t have either of these two reactions.

On one side of me, of course, are the sneering MAGA bullies who declare that might makes right, that the strong do what they can while the weak suffer what they must, and that the US should rule Venezuela for the same reason why Russia should rule Ukraine and China should rule Taiwan: namely, because the small countries have the misfortune of being in the large ones’ “spheres of influence.”

But on my other side are those who squeal that toppling a dictator, however odious, is against the rules, because right is whatever “international law” declares it to be—i.e., the “international law” that’s now been degraded by ideologues to the point of meaninglessness, the “international law” that typically sides with whichever terrorists and murderers have the floor of the UN General Assembly and that condemns persecuted minorities for defending themselves.

The trouble is, any given framework of law needs to do at least one of three things to impose its will on me:

1. Compel my obedience, by credibly threatening punishment if I defy it.
2. Win the assent of my conscience, by the force of its moral example.
3. Buy my consent through reciprocity: if this framework will defend my family from being murdered, I therefore ought to defend it.

But “international law,” as it exists today, fails spectacularly on all three of these counts. Ergo, as far as I’m concerned, it can take a long walk off a short pier.

Against these two attempted reductions of right to something that it isn’t, I simply say:

Right is right. Good is good. Evil is evil. Good is liberal democracy and the Enlightenment. Evil is authoritarianism and liars and bullies.

Good, in this case, is Maria Machado and the Venezuelans who went to prison, who took to the streets, who monitored every polling station to prove Edmundo González’s victory. Evil is those who oppose them.

But who gets to decide what’s good and what’s evil? Well, if you’re here asking me, then I decide.

But don’t the evildoers believe themselves to be good? Yes, but they’re wrong.

It’s crucial that I’m not appealing here to anything exotic or esoteric. I’m appealing only to the concepts of good and evil that I suspect every reader of this blog had as a child, that they got from fables and Disney movies and Saturday morning cartoons and the like, before some of them went to college and learned that those concepts were naïve and simplistic and only for stupid people.

Look: I regularly appear, to my amusement and chagrin, in Internet lists of the smartest people on earth, alongside Terry Tao and Garry Kasparov and Ed Witten. I did publish my first paper at 15, and finished my PhD in theoretical computer science at 22, and became an MIT professor soon afterward, yada yada.

And for whatever it’s worth, I’m telling you that I think the “naïve, simplistic” concepts of good and evil of post-WWII liberal democracy were fine all along, and not only for stupid people. In my humble opinion. Of course those concepts can be improved upon—indeed, criticism and improvement and self-correction are crucial parts of them—but they’re infinitely better than the realistic alternatives on offer from left and right, including kleptocracy, authoritarianism, and what we’re now calling “the warmth of collectivism.”

And according to these concepts, María Machado and the other Venezuelans who stand with her for democracy are good, if anything is good. Trump, despite all the evil in his heart and in his past, will do something profoundly good if he reverses himself and lets those Venezuelans have what they’ve fought for. He’ll do evil if he doesn’t.

Happy New Year, everyone. May goodness reign over the earth.