Monthly Archives: March 2018

Toward Reality TV MBAs

The quality of firm managers matters enormously for firm productivity. How can we get better managers? We already select the best people in terms of simple features like intelligence, conscientiousness, etc. But apparently there is still huge variation in quality, even after controlling for such things. Typical MBA programs teach people some business basics, but don’t seem to help much; they mainly serve to select elites and connect them to each other.

I recently had dinner with a few San Francisco tech startup CEOs, who were worth high sums. They weren’t obviously that much smarter etc. than others. Their high value came from having actually navigated difficult business waters, successfully enough. That sort of experience and track record is gold. Some said that business success came from making the right decision at a half dozen key points; any wrong move would have killed them.

Some had first gained experience via being a personal assistant to someone else in such a role. Such an assistant goes to all meetings and sees pretty much everything that manager does, over a several year period. Apparently children learn similar things via parents dinner conversations:

The majority of male entrepreneurs in Norway start a firm in an industry closely related to the one in which their father is employed. These entrepreneurs outperform others in the same industry. … ‘Dinner table human capital’ – that is, industry knowledge learned through their parents – is an important factor.… the effect of parents helping out, although possibly quite important, is smaller. (more; HT Alex T)

If one can learn much from just watching the inside story of real firms over several years, that suggests a big win: record the full lives of many rising managers over several years, and show a mildly compressed and annotated selection of such recordings to aspiring managers. Such recordings could be compressed by deleting sleep and non-social periods. They could be annotated to identify key decisions and ask viewers to make their own choices, before they see actual choices. Recordings might be selected 2/3 from the most successful, and 1/3 from a sampling of others.

Yes, there are issues of privacy and business secrets. But these are already issues for personal assistants and others who attend key business meetings. Waiting five years could take away many business secret concerns. And we don’t have to make these videos available to the world; making manager experiences visible to only 100 times more people might increase our pool of good manager candidates by a factor of 100. And that could be worth trillions to the world economy.

GD Star Rating
loading...
Tagged as: ,

Toward Micro-Likes

Long ago when electricity and phones were new, they were largely unregulated, and privately funded. But then as the tech (and especially the interfaces) stopped changing so fast, and showed big scale and network economies, regulation stepped in. Today social media still seems new. But as it hasn’t been changing as much lately, and it also shows large scale and network economies, many are talking now about heavier regulation. In this post, let me suggest that a lot more change is possible; we aren’t near the sort of stability that electricity and phones reached when they became heavily regulated.

Back in the early days of the web and internet people predicted many big radical changes. Yet few then mentioned social media, the application now most strongly associated with this new frontier. What did we miss? The usual story, which I find plausible, is that we missed just how much people love to get many frequent signals of their social connections: likes, retweets, etc. Social media gives us more frequent “attaboy” and “we see & like you” signals. People care more than we realized about the frequency, relative to the size, of such signals.

But if that’s the key lesson, social media should be able to move a lot further in this direction. For example, today Facebook has two billion monthly users and produces four million likes per minute, for an average of about three likes per day per monthly user. Twitter has 300 million monthly users, who send 500 million tweets per day, for less than two tweets per day per monthly user. (I can’t find stats on Twitter likes or retweets.) Which I’d say is actually a pretty low rate of positive feedback.

Imagine you had a wall-sized screen, full of social media items, and that while you browsed this wall the direction of your gaze was tracked continuously to see which items your gaze was on or near. From that info, one could give the authors or subjects of those items far more granular info on who is paying how much attention to them. Not only on how often how much your stuff is watched, but also on the mood and mental state of those watchers. If some of those items were continuous video feeds from other people, then those others could be producing many more social media items to which others could attend.

Also, so far we’ve usually just naively counted likes, retweets, etc., as if everyone counted the same. But we could instead use non-uniform weights based on popularity or other measures. And given how much people like to participate in synchronized rituals, we could also create and publicize statistics on what groups of people are how synchronized in their social media actions. And offer new tools to help them synchronize more finely.

My point here isn’t to predict or recommend specific changes for future social media. I’m instead just trying to make the point that a lot of room for improvement remains. Such gains might be delayed or prevented by heavy regulation.

GD Star Rating
loading...
Tagged as: , ,

The Uploaded

In this post I again contrast my analysis of future ems in Age of Em with a fictional depictions of ems, and find that science fiction isn’t very realistic, having other priorities. Today’s example: The Uploaded, by Ferrett Steinmetz:

The world is run from the afterlife, by the minds of those uploaded at the point of death. Living is just waiting to die… and maintaining the vast servers which support digital Heaven. For one orphan that just isn’t enough – he wants more for himself and his sister than a life of servitude. Turns out he’s not the only one who wants to change the world.

The story is set 500 years and 14 human generations after a single genius invented ems. While others quickly found ways to copy this tech, his version was overwhelming preferred. (In part due to revelations of “draconian” competitor plans.) So much so that he basically was able to set the rules of this new world, and to set them globally. He became an immortal em, and so still rules the world. His rules, and the basic tech and econ arrangement, have remained stable for those 500 years, during which there seems to have been vastly less tech change and economic growth than we’ve seen in the last 500 years.

His rules are the these: typically when a biological humans dies, one emulation of them is created who is entitled to eternal leisure in luxurious virtual realities. That one em runs at ordinary human speed, no other copies of it are allowed, ems never inhabit android physical bodies, and ems are never created of still living biological humans. By now there are 15 times as many ems as humans, and major decisions are made by vote, which ems always win. Ems vote to divert most resources to their servers, and so biological humans are poor, their world is run down, and diseases are killing them off.

Virtual realities are so engaging that em parents can’t even be bothered to check in on their young children now in orphanages. But a few ems get bored and want to do useful jobs, and they take all the nice desk jobs. Old ems are stuck in their ways and uncreative, preventing change. Biological humans are only needed to do physical jobs, which are boring and soul-crushing. It is illegal for them to do programming. Some ems also spend lots of time watching via surveillance cameras, so biological humans are watched all the time.

Every day every biological human’s brain is scanned and evaluated by a team of ems, and put into one of five status levels. Higher levels are given nicer positions and privileges, while the lowest levels are not allowed to become ems. Biological humans are repeatedly told they need to focus on pleasing their em bosses so they can get into em heaven someday. To say more, I must give spoilers; you are warned. Continue reading "The Uploaded" »

GD Star Rating
loading...
Tagged as: ,

How Deviant Recent AI Progress Lumpiness?

I seem to disagree with most people working on artificial intelligence (AI) risk. While with them I expect rapid change once AI is powerful enough to replace most all human workers, I expect this change to be spread across the world, not concentrated in one main localized AI system. The efforts of AI risk folks to design AI systems whose values won’t drift might stop global AI value drift if there is just one main AI system. But doing so in a world of many AI systems at similar abilities levels requires strong global governance of AI systems, which is a tall order anytime soon. Their continued focus on preventing single system drift suggests that they expect a single main AI system.

The main reason that I understand to expect relatively local AI progress is if AI progress is unusually lumpy, i.e., arriving in unusually fewer larger packages rather than in the usual many smaller packages. If one AI team finds a big lump, it might jump way ahead of the other teams.

However, we have a vast literature on the lumpiness of research and innovation more generally, which clearly says that usually most of the value in innovation is found in many small innovations. We have also so far seen this in computer science (CS) and AI. Even if there have been historical examples where much value was found in particular big innovations, such as nuclear weapons or the origin of humans.

Apparently many people associated with AI risk, including the star machine learning (ML) researchers that they often idolize, find it intuitively plausible that AI and ML progress is exceptionally lumpy. Such researchers often say, “My project is ‘huge’, and will soon do it all!” A decade ago my ex-co-blogger Eliezer Yudkowsky and I argued here on this blog about our differing estimates of AI progress lumpiness. He recently offered Alpha Go Zero as evidence of AI lumpiness:

I emphasize how all the mighty human edifice of Go knowledge … was entirely discarded by AlphaGo Zero with a subsequent performance improvement. … Sheer speed of capability gain should also be highlighted here. … you don’t even need self-improvement to get things that look like FOOM. … the situation with AlphaGo Zero looks nothing like the Hansonian hypothesis and a heck of a lot more like the Yudkowskian one.

I replied that, just as seeing an unusually large terror attack like 9-11 shouldn’t much change your estimate of the overall distribution of terror attacks, nor seeing one big earthquake change your estimate of the overall distribution of earthquakes, seeing one big AI research gain like AlphaGo Zero shouldn’t much change your estimate of the overall distribution of AI progress. (Seeing two big lumps in a row, however, would be stronger evidence.) In his recent podcast with Sam Harris, Eliezer said:

Y: I have claimed recently on facebook that now that we have seen Alpha Zero, Alpha Zero seems like strong evidence against Hanson’s thesis for how these things necessarily go very slow because they have to duplicate all the work done by human civilization and that’s hard. …

H: What’s the best version of his argument, and then why is he wrong?

Y: Nothing can prepare you for Robin Hanson! Ha ha ha. Well, the argument that Robin Hanson has given is that these systems are still immature and narrow, and things will change when they get general. And my reply has been something like, okay, what changes your mind short of the world actually ending. If your theory is wrong do we get to find out about that at all before the world does.

(Sam didn’t raise the subject in his recent podcast with me.)

In this post, let me give another example (beyond two big lumps in a row) of what could change my mind. I offer a clear observable indicator, for which data should have available now: deviant citation lumpiness in recent ML research. One standard measure of research impact is citations; bigger lumpier developments gain more citations that smaller ones. And it turns out that the lumpiness of citations is remarkably constant across research fields! See this March 3 paper in Science:

The citation distributions of papers published in the same discipline and year lie on the same curve for most disciplines, if the raw number of citations c of each paper is divided by the average number of citations c0 over all papers in that discipline and year. The dashed line is a lognormal fit. …

The probability of citing a paper grows with the number of citations that it has already collected. Such a model can be augmented with … decreasing the citation probability with the age of the paper, and a fitness parameter, unique to each paper, capturing the appeal of the work to the scientific community. Only a tiny fraction of papers deviate from the pattern described by such a model.

It seems to me quite reasonable to expect that fields where real research progress is lumpier would also display a lumpier distribution of citations. So if CS, AI, or ML research is much lumpier than in other areas, we should expect to see that in citation data. Even if your hypothesis is that only ML research is lumpier, and only in the last 5 years, we should still have enough citation data to see that. My expectation, of course, is that recent ML citation lumpiness is not much bigger than in most research fields through history.

Added 24Mar: You might save the hypothesis that research areas vary greatly in lumpiness by postulating that the number of citations of each research advance goes as the rank of the “size” of that advance, relative to its research area. The distribution of ranks is always the same, after all. But this would be a surprising outcome, and hence seems unlikely; I’d want to see clear evidence that the distribution of lumpiness of advances varies greatly across fields.

Added 27Mar: More directly relevant might be data on distributions of patent value and citations. Do these distributions vary by topic? Are CS/AI/ML distributed more unequally?

GD Star Rating
loading...
Tagged as: , ,

Growth Is Change. So Is Death.

The very readable book The Wizard and the Prophet tells the story of environmental prophet William Vogt investigating the apocalypse-level deaths of guano-making birds near Peru. When he discovered the cause in the El Nino weather cycle, his policy recommendations were to do nothing to mitigate this natural cause; he instead railed against many much smaller human influences, demanding their reversal. A few years later his classic 1948 screed Road To Survival, which contained pretty much all the standard environmental advice and concepts used today, continued to warn against any but small human-caused changes to the environment, while remaining largely indifferent to even huge natural changes.

I see the same pattern when people consider long term futures. People can be quite philosophical about the extinction of humanity, as long as this is due to natural causes. Every species dies; why should humans be different? And few get bothered by humans making modest small-scale short-term modifications to their own lives or environment. We are mostly okay with people using umbrellas when it rains, moving to new towns to take new jobs, etc., digging a flood ditch after our yard floods, and so on. And the net social effect of many small changes is technological progress, economic growth, new fashions, and new social attitudes, all of which we tend to endorse in the short run.

Even regarding big human-caused changes, most don’t worry if changes happen far enough in the future. Few actually care much about the future past the lives of people they’ll meet in their own life. But for changes that happen within someone’s time horizon of caring, the bigger that changes get, and the longer they are expected to last, the more that people worry. And when we get to huge changes, such as taking apart the sun, a population of trillions, lifetimes of millennia, massive genetic modification of humans, robots replacing people, a complete loss of privacy, or revolutions in social attitudes, few are blasé, and most are quite wary.

This differing attitude regarding small local changes versus large global changes makes sense for parameters that tend to revert back to a mean. Extreme values then do justify extra caution, while changes within the usual range don’t merit much notice, and can be safely left to local choice. But many parameters of our world do not mostly revert back to a mean. They drift long distances over long times, in hard to predict ways that can be reasonably modeled as a basic trend plus a random walk.

This different attitude can also make sense for parameters that have two or more very different causes of change, one which creates frequent small changes, and another which creates rare huge changes. (Or perhaps a continuum between such extremes.) If larger sudden changes tend to cause more problems, it can make sense to be more wary of them. However, for most parameters most change results from many small changes, and even then many are quite wary of this accumulating into big change.

For people with a sharp time horizon of caring, they should be more wary of long-drifting parameters the larger the changes that would happen within their horizon time. This perspective predicts that the people who are most wary of big future changes are those with the longest time horizons, and who more expect lumpier change processes. This prediction doesn’t seem to fit well with my experience, however.

Those who most worry about big long term changes usually seem okay with small short term changes. Even when they accept that most change is small and that it accumulates into big change. This seems incoherent to me. It seems like many other near versus far incoherences, like expecting things to be simpler when you are far away from them, and more complex when you are closer. You should either become more wary of short term changes, knowing that this is how big longer term change happens, or you should be more okay with big long term change, seeing that as the legitimate result of the small short term changes you accept.

But of course few are very good at resolving their near versus far incoherences. And so the positions people take end up depending a lot on how they first framed the key issues, as in terms of short or long term changes.

GD Star Rating
loading...
Tagged as: , ,

Study Resistance To Widened Political Polarization

Tyler Cowen today:

Consider how an economy might work if buying decisions were made on a consistently ideological basis. Imagine a “right-wing” supermarket chain and a “left-wing” alternative. … The history of Northern Ireland shows a great many retailers, from funeral parlors to bars, that served either a largely Protestant or a largely Catholic clientele. Maybe people felt better about these exclusive commercial affiliations, but it didn’t do the economy any favors to stifle competition, and it may have helped drive political polarization too.

Two days ago an economics professor mentioned to me that he was taking a class on how to mix drinks in part because that is a relatively unpoliticized sphere of life. While there are different drink philosophies, so far none have obtained strong political connotations. It seemed to him, and to me, that in many areas of life substantial fractions of people actively resist allowing different standard views there to collect political connotations.

Of course in a rising tide of polarization, more and more spheres of life may drown in political floods. Once major divisions within an area are seen as political, outside political allies may be drawn into a bitter fight, which one political side may win, enabling it to take over that area of life. But it is worth noticing that some social processes actively resist such widened polarization. (Or more precisely “pillarisation“.)

We would do well to study such processes. To identify which areas of life are now fighting how hard to resist being caught up in political polarization. Then to theorize on what causes this extra willingness to resist. Such theories may help resisting areas to better coordinate to resist polarization. Yes, many political groups are now organizing to infect more areas with political polarization. But there seems room for more coordination against such widened polarization. If only we understood at least the basics of what is going on here.

GD Star Rating
loading...
Tagged as:

The Model to Beat: Status Rank

There’s been much discussion of income inequality over the last few years. However, I just randomly came across what should be a seminal related result, published in 2010 but mostly ignored. Let me do my bit to fix that.

People often presume that policy can mostly ignore income inequality if key individual outcomes like health or happiness depend mainly on individual income. Yes, there may be some room for promoting insurance against income risk, but not much room. However, people often presume that policy should pay a lot more attention to inequality if individual outcomes depend more directly on the income of others, such as via envy or discouragement.

However, there’s a simple and plausible income interdependence scenario where inequality matters little for policy: when outcomes depend on rank. If individual outcomes are a function of each person’s percentile income rank, and if social welfare just adds up those individual outcomes, then income policy becomes irrelevant, because this social welfare sum is guaranteed to always add up to the same constant. Income-related policy may influence outcomes via other channels, but not via this channel. This applies whether the relevant rank is global, comparing each person to the entire world, or local, comparing each person only to a local community.

That 2010 paper, by Christopher Boyce, Gordon Brown, and Simon Moore, makes a strong case that in fact the outcome of life satisfaction depends on the incomes of others only via income rank. (Two followup papers find the same result for outcomes of psychological distress and nine measures of health.) They looked at 87,000 Brits, and found that while income rank strongly predicted outcomes, neither individual (log) income nor an average (log) income of their reference group predicted outcomes, after controlling for rank (and also for age, gender, education, marital status, children, housing ownership, labor-force status, and disabilities). These seem to me remarkably strong and robust results. (Confirmed here.)

The irrelevance of individual income and reference group income remained true whether the group within which a person was ranked was the entire sample, one of 19 geographic regions, one of 12 age groups, or one of six gender-schooling groups. This suggests that the actual relevant comparison group is relatively narrow. If people cared mainly about their global rank in the whole sample, then analyses of rank within groups should have missed an effect of the rank of the group, which should have appeared as an effect of reference group income. But such effects weren’t seen.

It these statistical models were the correct model of the world, then income policy could only include influence social welfare via the control variables of age, gender, education, marital status, children, housing ownership, labor-force status, and disabilities. You couldn’t improve social welfare directly by redistributing income, though redistribution or taxation might help by changing control variables.

But even that conclusion seems premature. The key idea here is that people care about their social status rank, and income should only be one of many factors contributing to social status. So we should really be looking at models where all of a person’s observable features can contribute to their status. For each feature, such as personality or marital status, we should ask if our data is best described as that factor contributing directly to social status, which is then ranked to produce individual outcomes, or whether that factor also influences individual outcomes via some other channel, that doesn’t pass through social status. It is only effects via those other channels that might change overall social welfare.

This seems a straightforward statistical exercise, at least for someone with access to relevant data. Who’s up for it?

GD Star Rating
loading...
Tagged as: , ,

Beware Covert War Morality Tales

For years I’ve been saying that fiction is mainly about norm affirmation:

Both religion and fiction serve to reassure our associates that we will be nice. In addition to letting us show we can do hard things, and that we are tied to associates by doing the same things, religious beliefs show we expect the not nice to be punished by supernatural powers, and our favorite fiction shows the sort of people we think are heroes and villains, how often they are revealed or get their due reward, and so on. (more)

People fear that story-less people have not internalized social norms well – they may be too aware of how easy it would be to get away with violations, and feel too little shame from trying. Thus in equilibrium, people are encouraged to consume stories, and to deludedly believe in a more just world, in order to be liked more by others. (more)

Our actual story abilities are tuned for the more specific case of contests, where the stories are about ourselves or our rivals, especially where either we or they are suspected of violating social norms. We might also be good at winning over audiences by impressing them and making them identify more with us, and we may also be eager to listen to gain exemplars, signal norms, and exert influence. (more) Continue reading "Beware Covert War Morality Tales" »

GD Star Rating
loading...
Tagged as: , ,