Monthly Archives: February 2019

Checkmate On Blackmail

Often in chess, at least among novices, one player doesn’t know that they’ve been checkmated. When the other player declares “checkmate”, this first player is surprised; that claim contradicts their intuitive impression of the board. So they have to check each of their possible moves, one by one, to see that none allow an escape.

The same thing sometimes happens in analysis of social policy. Many people intuitively want to support policy X, and they usually want to believe that this is due to the good practical consequences of X. But if the policy is simple enough, one may be able iterate through all the possible consequential arguments for X and find that they all fail. Or perhaps more realistically, iterate through hundreds of the most promising actual consequential arguments that have been publicly offered so far, and both find them all wanting, and find that almost all of them are repetitions, suggesting that few new arguments are to be found.

That is, it is sometimes possible with substantial effort to say that policy X has been checkmated, at least in terms of known consequentialist supporting arguments. Yes, many social policy chess boards are big, and so it can take a lot of time and expertise to check all the moves. But sometimes a person has done that checking on policy X, and then frequently encounters others who have not so checked. Many of these others will defend X, basically randomly sampling from the many failed arguments that have been offered so far.

In chess, when someone says “checkmate”, you tend to believe them, even if you have enough doubt that you still check. But in public debates on social policy, few people accept a claim of “checkmate”, as few such debates ever go into enough depth to go through all the possibilities. Typically many people are willing to argue for X, even if they haven’t studied in great detail the many arguments for and against X, and even when they know they are arguing with someone who has studied such detail. Because X just feels right. When such a supporter makes a particular argument, and is then shown how that doesn’t work, they usually just switch to another argument, and then repeat that process until the debate clock runs out. Which feels pretty frustrating to the person who has taken the time to see that X is in fact checkmated.

We need a better social process for together identifying such checkmated policies X. Perhaps a way that a person can claim such a checkmate status, be tested sufficiently thoroughly on that claim, and then win a reward if they are right, and lose a stake if they are wrong. I’d be willing to help to create such a process. Of course we could still keep policies X on our books; we’d just have to admit we don’t have good consequential arguments for them.

As an example, let me offer blackmail. I’ve posted seven times on this blog on the topic, and in one of my posts I review twenty related papers that I’d read. I’ve argued many times with people on the topic, and I consistently hear them repeat the same arguments, which all fail. So I’ll defend the claim that not only don’t we have good strong consequential arguments against blackmail, but that this fact can be clearly demonstrated to smart reasonable people willing to walk through all the previously offered arguments.

To review and clarify, blackmail is a threat that you might gossip about someone on a particular topic, if they don’t do something else you want. The usual context is that you are allowed to gossip or not on this topic, and if you just mention that you know something, they are allowed to offer to compensate you to keep quiet, and you are allowed to accept that offer. You just can’t be the person who makes the first offer. In almost all other cases where you are allowed to do or not do something, at your discretion, you are allowed to make and accept offers that compensate you for one of these choices. And if a deal is legal, it rarely matters who proposes the deal. Blackmail is a puzzling exception to these general rules.

Most ancient societies simply banned salacious gossip against elites, but modern societies have deviated and allowed gossip. People today already have substantial incentives to learn embarrassing secrets about associates, in order to gain social rewards from gossiping about those to others. Most people suffer substantial harm from such gossip; it makes them wary about who they let get close to them, and induces them to conform more to social pressures regarding acceptable behaviors.

For most people, the main effect of allowing blackmail is to mildly increase the incentives to learn embarrassing secrets, and to not behave in ways that result in such secrets. This small effect makes it pretty hard to argue that for gossip incentives the social gains out weigh the losses, but for the slightly stronger blackmail incentives, the losses out weight the gains. However, for elites these incentive increases are far stronger, making elite dislike plausibly the main consequentialist force pushing to keep blackmail illegal.

In a few recent twitter surveys, I found that respondents declared themselves against blackmail at a 3-1 rate, evenly split between consequential and other reasons for this position. However, they said blackmail should be legal in many particular cases I asked about, depending on what exactly you sought in exchange for your keeping someone’s secret. For example, they 12-1 supported getting your own secret kept, 3-2 getting someone to treat you fairly, and 1-1 getting help with child care in a medical crisis.

These survey results are pretty hard to square with consequential justifications, as the consequential harm from blackmail should mainly depend on the secrets being kept, not on the kind of compensation gained by the blackmailer. Which suggests that non-elite opposition to blackmail is mainly because blackmailers look like they have bad motives, not because of social consequences to others. This seems supported by the observation that women who trash each other’s reputations via gossip tend to consciously believe that they are acting helpfully, out of concern for their target.

As examples of weak arguments, Tyler Cowen just offered four. First, he says even if blackmail has good consequences, given current world opinion it would look bad to legalize it. (Never do the right thing if it looks bad?) Second, he says negotiating big important deals can be stressful. (Should all big deals be banned?) Third, it is bad to have social mechanisms (like gossip?) that help enforce common social norms on sex, gender and drugs, as those are mistaken. Fourth, making blackmail illegal somehow makes it easier for your immediate family to blackmail you, and that’s somehow better (both somehows are unexplained).

I’d say the fact that Tyler is pushed to such weak tortured arguments supports my checkmate claim: we don’t have good strong consequential arguments for making gossiper-initiated blackmail offers illegal, relative to making gossip illegal or allowing all offers.

GD Star Rating
Tagged as: ,

How Lumpy AI Services?

Long ago people like Marx and Engels predicted that the familiar capitalist economy would naturally lead to the immiseration of workers, huge wealth inequality, and a strong concentration of firms. Each industry would be dominated by a main monopolist, and these monsters would merge into a few big firms that basically run, and ruin, everything. (This is somewhat analogous to common expectations that military conflicts naturally result in one empire ruling the world.)

Many intellectuals and ordinary people found such views quite plausible then, and still do; these are the concerns most often voiced to justify redistribution and regulation. Wealth inequality is said to be bad for social and political health, and big firms are said to be bad for the economy, workers, and consumers, especially if they are not loyal to our nation, or if they coordinate behind the scenes.

Note that many people seem much less concerned about an economy full of small firms populated by people of nearly equal wealth. Actions seem more visible in such a world, and better constrained by competition. With a few big privately-coordinating firms, in contrast, who knows that they could get up to, and they seem to have so many possible ways to screw us. Many people either want these big firms broken up, or heavily constrained by presumed-friendly regulators.

In the area of AI risk, many express great concern that the world may be taken over by a few big powerful AGI (artificial general intelligence) agents with opaque beliefs and values, who might arise suddenly via a fast local “foom” self-improvement process centered on one initially small system. I’ve argued in the past that such sudden local foom seems unlikely because innovation is rarely that lumpy.

In a new book-length technical report, Reframing Superintelligence: Comprehensive AI Services as General Intelligence, Eric Drexler makes a somewhat similar anti-lumpiness argument. But he talks about task lumpiness, not innovation lumpiness. Powerful AI is safer if it is broken into many specific services, often supplied by separate firms. The task that each service achieves has a narrow enough scope that there’s little risk of it taking over the world and killing everyone in order to achieve that task. In particular, the service of being competent at a task is separate from the service of learning how to become competent at that task. In Drexler’s words: Continue reading "How Lumpy AI Services?" »

GD Star Rating
Tagged as: , ,

Conditional Harberger Tax Games

Baron Georges-Eugène Haussmann … transformed Paris with dazzling avenues, parks and other lasting renovations between 1853 and 1870. … Haussmann… resolved early on to pay generous compensation to [Paris] property owners, and he did. … [He] hoped to repay the larger loans he obtained from the private sector by capturing some of the increased value of properties lining along the roads he built. … [He] did confiscate properties on both sides of his new thoroughfares, and he had their edifices rebuilt. … Council of State … forced him to return these beautifully renovated properties to their original owners, who thus captured all of their increased value. (more)

In my last post I described abstractly how a system of conditional Harberger taxes (CHT) could help deal with zoning and other key city land use decisions. In this post, let me say a bit more about the behaviors I think we’d actually see in such a system. (I’m only considering here such taxes for land and property tied to land.)

First, I while many property owners would personally manage their official declared property values, many others would have them set by an agent or an app. Agents and apps may often come packaged with insurance against various things that can go wrong, such as losing one’s property.

Second, yes, under CHT, sometimes people would (be paid well to) lose their property. This would almost always be because someone else credibly demonstrated that they expect to gain more value from it. Even if owners strategically or mistakenly declare values too low, the feature I suggested of being able to buy back a property by paying a 1% premium would ensure that pricing errors don’t cause property misallocations. The highest value uses of land can change, and one of the big positive features of this system is that it makes the usage changes that should then result easier to achieve. In my mind that’s a feature, not a bug. Yes, owners could buy insurance against the risk of losing a property, though that needn’t result in getting their property back.

In the ancient world, it was common for people to keep the same marriage, home, neighbors, job, family, and religion for their entire life. In the modern world, in contrast, we expect many big changes during our lifetimes. While we can mostly count on family and religion remaining constant, we must accept bigger chances of change to marriages, neighbors, and jobs. Even our software environments change in ways we can’t control when new versions are issued. Renters today accept big risks of home changes, and even home “owners” face big risks due to job and financial risks. All of which seems normal and reasonable. Yes, a few people seem quite obsessed with wanting absolute guarantees on preservation of old property usage, but I can’t sympathize much with such fetishes for inefficient stasis. Continue reading "Conditional Harberger Tax Games" »

GD Star Rating
Tagged as: , ,