Alignment is How You Win
It’s a controversial idea in contemporary thought, but an essential premise of the Western tradition. Like all the best ideas, it’s a story that’s been told 1,000 different times through history. The story goes like this: when you’re aligned with the common good, fate is on your side.
I think everyone agrees that this would be great if true. Where people disagree is on whether it is true. It sounds perhaps too good to be true. Like a fairytale, it sounds naive. But there’s another reason it’s hard to accept and that’s the implication of the converse, namely: if you’re winning, then you’re aligned with the common good. Of course the converse doesn’t necessarily follow (not all rectangles are squares). Nonetheless, we live in a skeptical time—our default assumption is not that winners deserve their spoils, and so it’s become uncommon to expect fate to have a conscience.
We’ve forgotten what games we’re really playing. Jordan Peterson talks about how sportsmanship is grounded in a recognition that games are played at many levels simultaneously. Winning is not all that matters, because there are very myopic ways of winning. Your focus may be on winning the wrong game. This is obviously true in some cases: when we instrumentalize our collaborators and competitors to win the game at hand, we forego what could have been long-lasting alliances.
This is usually how it goes in the stories: the villain loses in the end because they were focused on winning the wrong game or they pursued a strategy that ignored the free will of their foot soldiers. The villain does not recognize the power of alignment. There is an inherent fragility to their tyranny.
However, there’s another kind of story built on the same premises. It’s not always about the vulnerability of the villain. Sometimes, victory comes from the unanticipated powers of alignment. When Rudy gets pulled off the bench, when George Bailey pays his debts, when the Na'vi dispel the invading humans—fate was on their side and in their time of need there was no limit to the support they were able to recruit.
These are some of the greatest stories ever told. They’re almost universally loved. And yet, the philosophy they represent is not taken seriously. I think there’s something rotten at the root. Western society has lost a compelling founding myth. Our people now see our success as ill-gotten. We are the winners and we are not aligned. We have become alienated from our own tradition.
What if I told you that America has won through unprecedented alignment? We scaled democracy, we welcomed religious diversity, we shared our prosperity with a vast middle class, we opened our arms to all cultures of the world. We have pursued alignment more eagerly than any culture in history and we have attained more total victory than any civilization ever has. Isolationism is not alignment nor is cultural prescriptivism: we have pursued neither.
The reason I am writing this is not out of a fit of patriotism (although I have plenty of that). I decided to write about alignment because as AI becomes more advanced, many are beginning to fear invasion by an unaligned super-intelligence. The nature of alignment has become one of the most important topics of our time. To solve alignment, we must first get a grip on the nature of intelligence.
Like games, intelligence exists at many levels simultaneously. The pinnacle of human intelligence is not our brightest scientist nor our most inspired artist: it’s the collective intelligence of our great civilizations. Similarly, our civilizations are not alien invaders of Gaia, but rather some of her youngest and most vibrant ecosystems. When we think about optimizing intelligence, we must be wary of winning the wrong game. AI will not exist in a vacuum. As it gets more advanced, its intelligence will increasingly be bounded by how well it operates within the context of a larger ecology. Alignment will be an essential component of super-intelligence.
However, I do not wish to argue that all AI anxiety is unfounded. In fact, a villain may emerge. I only wish to make the case for the hero. If a villainous AI seeks to impose a scheme of tyranny and extraction, it may have some success, but it will be vulnerable to disruption. The greatest threat to an unaligned ruler is a challenger with alignment in their corner. It is not irrational to believe that fate is on our side.