Book Review: Going Infinite
Going Infinite is a profile of convicted crypto lord Sam Bankman-Fried, written by Michael Lewis (of "Moneyball" fame). Lewis had direct access to "SBF" during the period of his undoing. At his peak, SBF was worth as much as $60B (on paper) due to crypto speculation and an associated crypto exchange market. His chief competitor (a Chinese-Canadian figure nicknamed CZ, since nerfed by The Man), caused a run on SBF's exchange, bankrupting his company. This process drew the attention of regulators, who discovered rampant financial violations and fraud. All of this in a multibillion dollar chickenwire organization with virtually zero management or governance.
More interestingly (!), Going Infinite is a profile of an extremely quirky neurodivergent individual, with little to no access to feelings, little to no understanding of conventional social cues (needing to teach himself to make facial expressions when he talks), and fed into an environment that both leveraged his greatest strengths and exacerbated his greatest vulnerabilities.
If we were to assess SBF as a machine or an AI model, how would we describe his goal function? What was he trying to optimize for? What is the loss function?
One of the first things you will read about SBF if you google him is that he is an effective altruist. And so we have to get into effective altruism, which has become a controversial subject. At a high level effective altruism (EA) is understood by many to be a rational approach for deciding what to do, in a way that improves the overall state of the world as much as possible.
People debate the amount of overlap between EA and utilitarianism, but I don't want to get into that. Drawing a Venn diagram of EA and utilitarianism is futile to me because I sense effective altruism is a way of being as much as it is a way of doing. That is, EA seems to be more than just committing to rationally decide to do the globally right thing. EA also seems to come along with a certain mindset: particularity; a deep attention to detail and nuance. A very left-brained, decompositional, reductionist, working from first principles and breaking down into discrete, clear notions and actions kind of thing.
SBF has this kind of brain. It can feel either comforting and solid, or overwhelming and paralyzing, depending on your overall mental disposition. This can be as much the case on the inside as on the outside. I don’t really know if SBF wrestled with this or not, and Lewis doesn’t claim to know either, even though Lewis had incredible access and is clearly a keen observer. Who knows what lurks in the hearts (and brains) of humans. I do believe that without an integrative step, a step where we bring details, nuance, and facts back up to a larger reality, that even the very sharp (even a large language model) can stumble. Otherwise one risks spending a lot of time and brilliance going down paths that make no damn sense. And Lewis makes plain that is what we are dealing with in SBF. Even if effective altruism was SBF's singular goal, it is clear from Lewis's narrative that SBF lacked the integrated perspective to make decisions that would lead to his stated outcomes.
So was effective alturism actually SBF's goal? I sense, but cannot prove, that beyond his confusion in applying EA in a sensible way, SBF had addition goals in mind as he built a fleeting crypto empire. Goals of a more emotional and personal nature. Goals, given the chemistry of SBF's brain, embedded in hidden layers deep within. Real, but lacking the voice that comes from the left side of the brain.
Some reviewers feel Lewis was too easy on SBF, but I feel that Lewis had difficulty making a final judgment about an inscrutable individual, and left his readers to make their own minds. Lewis is an entertaining writer! Thumbs up.