Robin Hanson - The Long View

The interview opens with an obvious objection to Robin Hanson's essays on the long view: if almost no significant player optimizes for the long term, perhaps that is because the future is too unpredictable, or because attention to it distracts from present circumstances. Hanson's answer is that unpredictability is itself something an optimizer should price in. The puzzle is that people appear to deviate from what they would do if they were optimally accounting for the long term and for the real amount of uncertainty.

His main evidence is the discount rate. Someone with the long view would use a very low discount rate and treat the future as nearly as important as the present. He adds that uncertainty does not make this hopeless, because resources are generally useful whatever happens. You need not know exactly what the future holds to know that collected resources will help. Discounting the future while declining to accumulate resources is therefore a clear sign that a person or institution is not much interested in the long-term future.

The host asks why, if long-term thinking is useful, today's most powerful institutions are not those that thought about the long term in the past. Hanson says this is part of the question he raises: over a very long run one might expect selection in favor of long-term thinking. Biology is his model. Species and individuals compete to influence the future, and having more descendants is the main way biology does so.

Planning inside this competition carries a built-in discount rate. Individual genes cannot save resources for themselves, only for the whole creature, which is a bundle of genes. A creature that saves resources for its descendants passes them to offspring who share only half of its genes. A gene therefore chooses between promoting itself now and collecting resources that give it a 50 50 chance of being promoted later. This produces a discount factor of two per generation, which Hanson says is in the ballpark of actual human discount rates. He offers it as a plausible explanation for those rates: humans evolved in a setting where saving for the next generation benefited a group sharing only half their genes.

Hanson frames the challenge as finding a unit of selection that can pass resources to itself over the long run. Candidates include a nation, a firm, a religion or another organizational unit, and he expects that in the long run organizations will find a way to collect resources and promote themselves. So far they have not, because organizations have little separate existence in our world. They are tools that genes, individuals and organizations use for their own ends. If an organization announced that it cared only about promoting itself, the individuals and genes would rebel and bring it back in line.

Religions that encourage proselytizing come closest, because they found a win-win arrangement in which individuals promote themselves by promoting the religion. Such deals are hard to find. Accumulating resources is the harder problem, since once a church, firm or club has resources, the humans in charge are tempted to grab them for themselves and their genes. The open question is how an organization can collect resources and keep them directed at the organization's purpose rather than at those who run it.

Asked whether he has considered raising money to build such an institution, Hanson says he has thought it through only to understand what is going on, and that he is not rich. He explains that contracts could in principle solve the problem. An organization could have a constitution that members must accept, enforced by law, specifying that resources be saved, and the organization would not be taxed or destroyed. According to Hanson, what went wrong historically is that people associated with such organizations went to the courts, wanting those resources, and pushed judges to interpret the law so that these contracts were not enforced. He concludes that law has specifically prevented arbitrary organization contracts that let an organization save resources into the future.

The host suggests that a nation state fits the description of an evolutionary agent that grows more powerful. Hanson says it does not, because the people running a nation state get to spend its resources. In a democracy, voters see accumulated resources and elect politicians who will spend them on the voters. Even a dictator sets things up for himself or perhaps his children, not for the organization to keep collecting resources forever.

Hanson then walks through the economics. For a very long time, as far as is known, rates of return on assets have exceeded economic growth rates. His example is an economy that doubles in 20 years while assets double in 10. Any organization that chooses to grow its assets would grow as a fraction of the economy until it dominated the whole economy. At that point interest rates would fall until the organization could not earn a return above the growth rate.

This opportunity has been available for a long time without being exploited. Nobody has managed to create organizations that keep lasting and collecting resources while growing faster than the economy. Either the people running the organization decide to take the resources, or outsiders do: a local city or government taxes or grabs them, or invaders take them in war. Hanson still predicts that eventually some unit of selection will take the long view, collect resources, make choices and promote itself with the long run in mind. When enough such units exist, they will push interest rates down so that they do not exceed growth rates.

The implication he draws concerns neglected risks. Today one may fear that things will happen in the long run that could destroy us, yet nobody is much interested in working to prevent them, because there are no units that take the long view. Individual humans do not, because of their discount rates. Organizations do not, because they are run by humans, and other biological species cannot plan or think about the long run. Once the world is full of long-view creatures, the long term would no longer be neglected.

The host asks whether religion, which implicitly acts as if it plans for the long term in some ways, shows the process at work. Hanson says organizations are not autonomous agents that plan everything. They are collections of strategies that humans use, and in some ways they manage to promote themselves and in others they do not. They are not conscious planners with a utility function who can make complicated plans.

The host replies that this is what one would expect in an evolutionary landscape, since eventually unconscious, memetic cultural units might succeed. Hanson answers on two points. First, organizations are new. For about a million years human organizations were bands of around 50 people, a distinctive achievement compared with other primates that took a lot of work. With farming 10,000 years ago, villages of around a thousand people became possible, and then empires, but empires controlled a limited range of what people did. If an empire came by wanting soldiers and taxes, villagers supplied them, and otherwise the empire did not run things.

Second, predicting that a unit will plan successfully to promote itself requires a unit that can plan. A meme is a tiny piece that cannot promote itself except together with many others that form organizations. Genes work the same way: they promote themselves through organisms, which are large complicated collections of genes, because an individual gene cannot plan, store resources or make strategies. Organizations likewise need many parts that all work together, and whether that can be achieved is the open question.

The host then cites Tyler Cowen's argument in Stubborn Attachments, as transcribed, that someone who cares about the long term should simply promote economic growth. Hanson says it depends on what you care about in the long run. If you care only that everybody is happy, regardless of whether you are allied with them or like them, you want the future to be big, and economic growth is by definition a measure of how much stuff is happening. His own prediction of organizations with a long view is a different matter. It does not assume that preference, which he says is not stable under competition. Biological organisms take a long view in order to promote themselves and create descendants, which is not the same as wanting the whole earth to have more biomass.