Wisdom Of The Week

Lesson 1 - Scope Matters

In discussing large-scale models, it’s difficult to avoid mentioning Jorge Luis Borges’s thought experiment about a 1:1 scale map from “On Exactitude in Science”:

“In time, . . . the cartographers guilds struck a map of the empire whose size was that of the empire, and which coincided point for point with it. The following generations, who were not so fond of the study of cartography as their forebears had been, saw that that vast map was useless, and not without some pitilessness was it, that they delivered it up to the inclemencies of sun and winters. In the deserts of the west, there are tattered ruins of that map, inhabited by animals and beggars, in all the land there is no other relic of the disciplines of geography.”

The point we take from Borges (and from Cheramie) is that no model can be a complete recapitulation of the real world. Instead, we bracket off parts of the world, model those parts, and use the insights it gives us to make interventions in the world. The Army Corps couldn’t model the entire Missippi Basin drainage system either. They could only follow tributaries so far upstream before having to make generalized assumptions about the inputs to the system they modeled. They also couldn’t model all the outputs - their model doesn’t extend past Baton Rouge, let alone out into the Gulf of Mexico.

Similarly, the inputs for computer models are the outputs of other processes not captured by the model itself, and so the outputs of a model are only as valid as the understanding of the conditions that feed into it. If a minor creek jumps its bank upstream from region modeled by the Mississippi Basin Model, it could have downstream effects that the model could never capture. If the conditions that produce the data points we use to feed our model change, so too can the validity of our model change. The success of projects like AlphaGo rely on modeling closed systems, e.g. the game of Go, which is why AI for games are (relatively) easy and applied, real-world AI is much harder. Machine learning is great at predicting the future when the future resembles the past, but it takes a lot more to predict the lay of the land when the ground shifts under our feet.

Lesson 2 - Materials Matter

In building their Mississippi Basin Model, the Army Corps had to approximate the “real world” with the materials they had at their disposal. The engineers shaped and textured concrete, installed brass plugs, and accordion-folded sheet metal to approximate the incredibly complex effects of trees, sand, clay, roads, and crops on the speed, direction, and volume of water passing over the landscape in high-water conditions. They had to develop a measure of “frictional resistance” to translate between the real world of rocks and trees and the model world of concrete and metal. In computer modeling, the proxies we choose to represent the real world are just as important. We don’t know where people are, necessarily, but we do have a great degree of confidence about where their GPS-enabled phones are. Similarly, another example of this comes from the world of computer vision, where attempts to produce soccer highlights from video struggled with following the ball (exciting moments are more likely the closer the ball is to the goal). Eventually, one team discovered that players tend to follow the ball, and players are easier to track, so the players became a useful proxy for addressing a harder question.

It is from these approximations of reality that we’re able to train the coefficients of our models, and so, importantly, the proxies we choose are the materials that shape how inputs relate to outputs. The models themselves have a material affect on outputs, too. If we assume that inputs are linear, and put them in to a linear model, they will produce a linear output. If the relationship between inputs and outputs is not actually linear, then the model will not fit, in every sense of the word. The Mississippi Basin Model had to pick and choose what it could approximate, and reduce everything else to coefficients. Wetlands disappeared form the model, as did evaporation and siltation. The lesson Cheramie draws from this is that “it doesn’t matter how much territory the model covers if it relies on the amputation of inconvenient complexities to be manageable. The simulation becomes thin.” Computer models can manage a great deal more complexity than physical models, but the crucial complexity that data scientists should pay careful attention to is the material relationship between the reality we hope to model and the proxies we choose to represent that reality. Neural networks with external memory, that learn to remember and recollect, are attempts to build “context awareness” and long-term memory into neural networks. This can be understood as an attempt to model a larger chunk of the world, to bring in more materials without having to explicitly declare every variable worth considering.

- Learning from Real-World Models: The Mississippi Basin Model and Machine Learning

Quote of the Day

The most exciting phrase to hear in science, the one that heralds new discoveries, is not ‘Eureka!’ but ‘That’s funny…’

- Isaac Asimov

Quote of the Day

Quote of the Day

I have seen many storms in my life. Most storms have caught me by surprise, so I had to learn very quickly to look further and understand that I am not capable of controlling the weather, to exercise the art of patience and to respect the fury of nature.

- Paulo Coelho

Quote of the Day

Medicine is a science of uncertainty and an art of probability.

- Sir William Osler

Quote of the Day

Of all the plants that cover the earth and lie like a fringe of hair upon the body of our grandmother, try to obtain knowledge that you may be strengthened in life.

- Winnebago (Native American) (on nature)

Quote of the Day

The contemporary proliferation of bullshit also has deeper sources, in various forms of skepticism which deny that we can have any reliable access to an objective reality and which therefore reject the possibility of knowing how things truly are. These "anti-realist" doctrines undermine confidence in the value of disinterested efforts to determine what is true and what is false, and even in the intelligibility of the notion of objective inquiry. One response to this loss of confidence has been a retreat from the discipline required by dedication to the ideal of correctness to a quite different sort of discipline, which is imposed by pursuit of an alternative ideal of sincerity. Rather than seeking primarily to arrive at accurate representations of a common world, the individual turns toward trying to provide honest representations of himself. Convinced that reality has no inherent nature, which he might hope to identify as the truth about things, he devotes himself to being true to his own nature. It is as though he decides that since it makes no sense to try to be true to the facts, he must therefore try instead to be true to himself.

But it is preposterous to imagine that we ourselves are determinate, and hence susceptible both to correct and to incorrect descriptions, while supposing that the ascription of determinacy to anything else has been exposed as a mistake. As conscious beings, we exist only in response to other things, and we cannot know ourselves at all without knowing them. Moreover, there is nothing in theory, and certainly nothing in experience, to support the extraordinary judgment that it is the truth about himself that is the easiest for a person to know. Facts about ourselves are not peculiarly solid and resistant to skeptical dissolution. Our natures are, indeed, elusively insubstantial -- notoriously less stable and less inherent than the natures of other things. And insofar as this is the case, sincerity itself is bullshit.

- Harry G. Frankfurt, On Bullshit

John Roberts’ Commencement Speech @ his Son’s 9th Grade Graduation

Now the commencement speakers will typically also wish you good luck and extend good wishes to you. I will not do that, and I’ll tell you why. From time to time in the years to come, I hope you will be treated unfairly, so that you will come to know the value of justice. I hope that you will suffer betrayal because that will teach you the importance of loyalty. Sorry to say, but I hope you will be lonely from time to time so that you don’t take friends for granted. I wish you bad luck, again, from time to time so that you will be conscious of the role of chance in life and understand that your success is not completely deserved and that the failure of others is not completely deserved either. And when you lose, as you will from time to time, I hope every now and then, your opponent will gloat over your failure. It is a way for you to understand the importance of sportsmanship. I hope you’ll be ignored so you know the importance of listening to others, and I hope you will have just enough pain to learn compassion.

Whether I wish these things or not, they’re going to happen. And whether you benefit from them or not will depend upon your ability to see the message in your misfortunes.

- via MR

Quote of the Day

Don't wait for the Last Judgement. It takes place every day.

- Albert Camus

Wisdom Of The Week

The METI group aims to improve on the Arecibo message not just by targeting specific planets, like that super-earth orbiting Gliese, but also by rethinking the nature of the message itself. ‘‘Drake’s original design plays into the bias that vision is universal among intelligent life,’’ Vakoch told me. Visual diagrams — whether formed through semiprime grids or engraved on plaques — seem like a compelling way to encode information to us because humans happen to have evolved an unusually acute sense of vision. But perhaps the aliens followed a different evolutionary path and found their way to a technologically advanced civilization with an intelligence that was rooted in some other sense: hearing, for example, or some other way of perceiving the world around them for which there is no earthly equivalent.

Like so much of the SETI/METI debate, the question of visual messaging quickly spirals out into a deeper meditation, in this instance on the connection between intelligence and visual acuity. It is no accident that eyes developed independently so many times over the course of evolution on Earth, given the fact that light conveys information faster than any other conduit. That transmission-speed advantage would presumably apply on other planets in the Goldilocks zone, even if they happened to be on the other side of the Milky Way, and so it seems plausible that intelligent creatures would evolve some sort of visual system as well.

But even more universal than sight would be the experience of time. Hans Freudenthal’s ‘‘Lincos: Design of a Language for Cosmic Intercourse,’’ a seminal book of exosemiotics published more than a half-century ago, relied heavily on temporal cues in its primer stage. Vakoch and his collaborators have been working with Freudenthal’s language in their early drafts for the message. In Lincos, duration is used as a key building block. A pulse that lasts for a certain stretch (say, in human terms, one second) is followed by a sequence of pulses that signify the ‘‘word’’ for one; a pulse that lasts for six seconds is followed by the word for six. The words for basic math properties can be conveyed by combining pulses of different lengths. You might demonstrate the property of addition by sending the word for ‘‘three’’ and ‘‘six’’ and then sending a pulse that lasts for nine seconds. ‘‘It’s a way of being able to point at objects when you don’t have anything right in front of you,’’ Vakoch explains.

[---]

All of which takes us back to a much more down-to-earth, but no less challenging, question: Who gets to decide? After many years of debate, the SETI community established an agreed-­upon procedure that scientists and government agencies should follow in the event that the SETI searches actually stumble upon an intelligible signal from space. The protocols specifically ordain that ‘‘no response to a signal or other evidence of extraterrestrial intelligence should be sent until appropriate international consultations have taken place.’’ But an equivalent set of guidelines does not yet exist to govern our own interstellar outreach.

One of the most thoughtful participants in the METI debate, Kathryn Denning, an anthropologist at York University in Toronto, has argued that our decisions about extraterrestrial contact are ultimately more political than scientific. ‘‘If I had to take a position, I’d say that broad consultation regarding METI is essential, and so I greatly respect the efforts in that direction,’’ Denning says. ‘‘But no matter how much consultation there is, it’s inevitable that there will be significant disagreement about the advisability of transmitting, and I don’t think this is the sort of thing where a simple majority vote or even supermajority should carry the day . . . so this keeps bringing us back to the same key question: Is it O.K. for some people to transmit messages at significant power when other people don’t want them to?’’

In a sense, the METI debate runs parallel to other existential decisions that we will be confronting in the coming decades, as our technological and scientific powers increase. Should we create superintelligent machines that exceed our own intellectual capabilities by such a wide margin that we cease to understand how their intelligence works? Should we ‘‘cure’’ death, as many technologists are proposing? Like METI, these are potentially among the most momentous decisions human beings will ever make, and yet the number of people actively participating in those decisions — or even aware such decisions are being made — is minuscule.

‘‘I think we need to rethink the message process so that we are sending a series of increasingly inclusive messages,’’ Vakoch says. ‘‘Any message that we initially send would be too narrow, too incomplete. But that’s O.K. Instead, what we should be doing is thinking about how to make the next round of messages better and more inclusive. We ideally want a way to incorporate both technical expertise — people who have been thinking about these issues from a range of different disciplines — and also getting lay input. I think it’s often been one or the other. One way we can get lay input in a way that makes a difference in terms of message content is to survey people about what sorts of things they would want to say. It’s important to see what the general themes are that people would want to say and then translate those into a Lincos-like message.’’

When I asked Denning where she stands on the METI issue, she told me: ‘‘I have to answer that question with a question: Why are you asking me? Why should my opinion matter more than that of a 6-year-old girl in Namibia? We both have exactly the same amount at stake, arguably, she more than I, since the odds of being dead before any consequences of transmission occur are probably a bit higher for me, assuming she has access to clean water and decent health care and isn’t killed far too young in war.’’ She continued: ‘‘I think the METI debate may be one of those rare topics where scientific knowledge is highly relevant to the discussion, but its connection to obvious policy is tenuous at best, because in the final analysis, it’s all about how much risk the people of Earth are willing to tolerate. . . . And why exactly should astronomers, cosmologists, physicists, anthropologists, psychologists, sociologists, biologists, sci-fi authors or anyone else (in no particular order), get to decide what those tolerances should be?’’

- Greetings, E.T. (Please Don’t Murder Us.) by Steven Johnson