Seth Goldstein

I am slowly starting to settle into a routine here in California.  The past few months have been filled with new beginnings- a new school for the kids, a new job and commute for Tina, new grocery stores and restaurants and little league fields.  The Taxi cab-hailing hustle of Manhattan has given way to hustling my bike to the top of Mt Tam.

Ideas don’t change at the same pace as activities, however, and I find myself thinking through the same issues about transparency that stimulated this blog in the first place.  Back then I was focused on soft dollars and the opacity of financial markets not media markets:

…soft dollars and bundled commissions are the vig that generates much of the wealth among the brokerage industry in New York, which in turn lubricates expense accounts at lunch time and grand Park Avenue co-ops and East Hampton beachfronts. Is it not ironic that New York has a mayor whose namesake company benefits more from monthly soft dollar payments than perhaps any other financial institution. In a way, Bloomberg has taken the notion of value-added brokerage services to the peak of civic duty. Our city itself reflects the residual value of opacity in financial markets. And so the question comes back to what happens to the brokerage industry when transparency become of more value to investors than opacity?

Three years later, soft dollar pratices are as opaque as ever.  The SEC has catered to the rich interests of hedge fund and stock brokerage lobbyists and enabled both sides to continue their practice of doing business with eachother in a very gray market.  Even the most sophisticated individuals outside of the financial services industry have little sense of what is really going on, in terms of the ways in which large institutional investors and large banks and brokers profit from closed data practices.

This is not that different from the dynamics of the online advertising environment- there are large institutional advertisers doing business with large media companies and advertising networks.  Despite the pre-text of openness and transparency, the online media market works hard to obscure the discovery of price by the very individuals producing it; namely the people who are using the medium, searching for things, clicking on ads, and conducting commercial transactions.  The web user, like the individual  investor, has resigned himself to letting larger interests capture, aggregate, and monetize his data behavior.  He has been led to believe that this is simply part of the bargain of having such "low" transaction costs for trading stocks or searching for information.

If I had to draw a continuous line through all of my disparate activities over the past ten years, this would be it:  identifying and interpreting the direct economic value of an individual data actor.  We may never in our lifetime see a day when a person develops an acute, vested interest in the value of his data; the spread between the value of a handful of clicks and that of a mass of aggregate behavior is significant.

Although it may be hard to keep track of the progression of Media Futures, we are stuck between the end of Alchemy and the beginning of Arbitrage.  This is where the creativity stops and the money kicks in; not that surprising against the backdrop of so much M&A activity (DoubleClick, RightMedia, StumbleUpon, etc.)

In May 2005, I made this transition in the first Media Futures series.  It was etymological in nature and only hinted at the real activities that I was engaged with as an entrepreneur, as I handed over the reins of Majestic Research to a new CEO in order to focus on creating Root Markets.  Flash forward two years and I am at a similar juncture; this time moving from Root in NY to AttentionSoft in SF.

The transition from Alchemy to Arbitrage that I want to describe this time will be more personal, now that the philosophical ground work has been established.  I want to trace the evolution of a central idea- transparency- through the founding of a new investment research process in 2002 all the way through the creation of a new consumer data platform in 2007.

As always, thanks for staying tuned.

Ever wonder how algorithms threaten our identities?

Tune into AttentionGate, the new series from the folks that brought you Identity Theft

In the Arrival, the first episode in 1967 of the British TV show The Prisoner, Number 6 says:

“I will not make any deals with you. I’ve resigned. I will not be pushed, filed, stamped, indexed, briefed, debriefed or numbered. My life is my own.” 

If an algorithm is an operation that turns a certain input into an equally certain output, then it can be seen as expressing a unique and consistent identity.  The extent to which this algorithm, in trying to faithfully represent a pre-existing identity, in the process reduces that identity to something less than itself, then the algorithm no longer represents identity.  Instead, the algorithm disturbs, distorts and destroys the identity.  It becomes a new, different identity- maybe still disguised as the underlying, organic, authentic, original human identity but actually nothing more than a bionic simulation. 

When the input for an algorithm is the attention of the user, the success of the algorithm is based on its ability to record attention data at its original resolution, preserved in its original context.  For example, it’s not only the specific area of the page I was paying attention to, but who influenced me to go there in the first place.  The failure of attention algorithms to maintain this fidelity began with “My Tivo Thinks I am Gay”

In trying to correct the machine from thinking he was gay, Mr. Iwanyk “tried to tame TiVo’s gay fixation by recording war movies and other “guy stuff.”  His attempt to recalibrate the attention algorithm is too strong, and its starts to generate “documentaries on Joseph Goebbels and Adolf Eichmann” as output.

The specter of the Holocaust casts its shadow once again.  Data costs merge into datacaust.  Before we had identity theft in the information world, we had
identity theft in the physical world.  This was not about stealing a
virtual identity, but about stealing physical bodies.  Removing human
expression (hairstyle, clothing, jewelry, social circle) and reducing
the individual to a single number was the best way to eliminate human identity without eliminating human productivity.

As the number of prisoners brought to the expanding Auschwitz
complex rose, so did the death rate. But if a corpse were separated
from its uniform, identification was rendered all but impossible. With
often hundreds of prisoners dying per day, other methods of
identification were needed… In May 1944, numbers in the "A" series
and the "B" series were first issued to  Jewish prisoners, beginning
with the men on May 13th and the women on May 16th. The "A" series was
to be completed with 20,000; however an error led to the women being
numbered to 25,378 before the "B" series was begun. The intention was
to work through the entire alphabet with 20,000 numbers being issued in
each letter series. In each series, men and women had their own
separate numerical series, ostensibly beginning with number 1…  from The Center for Holocaust and Genocide Studies

 
Today, the seductiveness of Tivo has been eclipsed by the sensitive ears of search boxes.  Every interest is heard, every desire is remembered, every curiosity recorded on somebody else’s server.  In exchange for capturing all of this information, the search algorithm promises to return relevant links that deliver a profitable return on attention. 

When we wonder about the black box, and ask to see the wizard inside of an attention algorithm, invariably that black box opens up to reveal another one: at the same moment link-based algorithms are losing the battle against splogs and fraudulent clicks, personalized search emerges and seduces us effortlessly into a new attention algorithm; at least until people start wondering about the value of consuming their own personal attention residues and log out out from the platform.

Stay tuned for upcoming AttentionGate episodes on AOL search data and pretext surveillance at HP …

An algorithm is a machine that can be used to reproduce a unique pattern of behavior.   The history of the word traces back to the Greeks and the instruments they used for mathematics; for example, the sieve.  In the context of Media Futures, imagine that algorithms are tightly woven filters that capture the full range of human Automata and slowly sift through them to produce the most meaningful, intentional gestures.

Ancient Algorithms

Finding its root in algorism, a reading of the name of Abu Ja’far Muhammad ibn Musa Al-Khwarizmi, the 9th century Persian mathematician who described a set of rules for solving both Linear and Quadratic equations, algorithm came to its present state by way of an 18th century European Latin translation and soon expanded its meaning to encompass all definite procedures for solving problems or performing tasks.  The very first algorithms are a part of the Babylonian mathematical legacy – a legacy which not only left us with algorithms for factorization, finding square roots and performing long division, but which also left us with the base 60 system that gives 60 minutes to an hour, 60 seconds to a minute, 360 degrees to a circle and 24 hours to a clock.  Babylonians were in fact able to calculate things with the same accuracy as Renaissance mathematicians due to their use of number tables, like the Plimpton Tablet, a table of Pythagorean Triples from about 1700 B.C.

While the Babylonians based their mathematical system in large part on algebra, the Greek system of mathematics was heavily based upon geometry.  It is speculated, though, that the founder of Greek science and mathematics, the philosopher Thales of Milet, visited Egypt and Babylon during his lifetime (634 – 546 B.C.) and brought back knowledge of their astronomy and geometry.  The Egyptians made great contributions in the fields of medicine, astronomy and applied mathematics, and while the former triumphs are well documented, there exist no records of the process by which they reached their mathematical conclusions.  Thales built on the knowledge brought back from his trips, inventing deductive mathematics and proving a number of theorems – a circle is bisected by a diameter; the base angles of an isosceles triangle are equal; and pairs of vertical angles formed by two intersecting lines are equal.

The foremost text on geometry came from fellow Greek Euclid, whose Elements put together former geometric knowledge with definitions, postulates and opinions – and, of course, Euclid’s elegant and rigorous proofs of the above.  In that text, he discussed the algorithm for finding the greatest common divisor of two numbers, which is today referred to as the Euclidean algorithm.  One hundred years later around 200 B.C., the world saw the next great algorithm – the Sieve of Eratosthenes, which was used to find prime numbers.

From Wikipedia: 

Sieve of Eratosthenes is a simple, ancient algorithm for finding all prime numbers up to a specified integer. It is the predecessor to the modern Sieve of Atkin, which is faster but more complex. It was created by Eratosthenes, an ancient Greek mathematician.

Another important site in the history of the algorithm was Alexandria, home to Hero, Ptolemy, and Diophantos.  Hero, whom we will remember as the inventor of the steam eolipile and other Automata, published widely on geometrics, optics and mechanics – as well as mathematics.  Though sources suggest his work is derivative of Archimedes and the work of the Babylonians, his Formula to calculate the area of a triangle in terms of its sides and his Method to extract a root are important contributions to the world of mathematics.  Ptolemy published widely on astronomy and geography and calculated the best approximation of ‘pi’ for his time.  And Diophantos, known as the ‘father of algebra’, wrote his thirteen-volume Arithmetica on the solution of algebraic equations and the theory of numbers and introduced the use of algebraic symbolism with an abbreviation for the unknown for which he was solving.

But Diophantos shares the title of the ‘father of algebra’ with the aforementioned Al-Khwarizmi, whose work was responsible for significant advances in the world of mathematics. 

It was Al-Khwarizmi’s work that promoted the use of Hindu-Arabic numerals that not only pushed forward the numeral system we use today, but that gives us the very term algorithm. From the very first algorithms of the Babylonians to those of Al-Khwarizmi – to John Napier’s 1614 method for performing calculations using logarithms to the 19th century work of Boole, Frege and Peano, which set out to reduce arithmetic to a series of symbols which could be manipulated by rules – to the work of Babbage, Lovelace and Turing, which took these rules and transformed them into agents of action in computing, these feats of problem-solving are instrumental in understanding man’s quest for a grasp of the workings of the world at large.      

Babbage and Turing

One great advantage which we may derive from machinery
is from the check which it affords against the inattention, the
idleness, or the dishonesty of human agents.
From Babbage’s 1832 work “On the Economy of Machinery and Manufactures”

In our discussion of rules that govern the Internet, we must turn to the work of Babbage and Turing, for it serves as the important foundation for computing at all.  Babbage’s work grew out in part out of a need for more accurate mathematical tables, which were essential calculating aids used in navigation and astronomy, insurance and civil engineering.  These tables were produced by human computers and by hand – and as such, they were prone to error in terms of computation and reporting.  Even the slightest errors in navigational or astronomical tables can be costly – so it is no surprise that in the years leading up to Babbage’s project, government sources were willing to fund projects that would minimize the costs of troubleshooting. 

For example, the British Nautical Almanac, the world’s first permanent table-making project – had a reputation for ever-improving accuracy since its inception in 1766.  But moving into the 19th century, that seaman’s bible swung into a dangerous territory of inaccuracy and error, and the British government recognized the promise of producing mathematical tables mechanically and typesetting them by the same machine. 

So Babbage set out, with financial support (and the admirals’ prayers) to improve the accuracy of those ever-important mathematical tables by constructing algorithm-driven machines.  It was a move that mechanized the production of thought, a move that would eliminate human folly in computation, transcription and typesetting.  The result would be better answers, answers which would in turn be used for giving new instructions, as inputs in other algorithms.   

Babbage never finished his Difference Engine – though, in 1832 his manufacturing engineer did construct a working portion of it, which measured two and a half feet high by two feet wide by two feet deep.  Babbage moved forward to conceptualizing what would be the world’s first programmable digital computer – the Analytical Engine.  Babbage’s designed the engine such that it would separate the sites of arithmetic computation from the storage of numbers.  The computation would be carried out through a series of steps recorded on punch cards, such as the ones used in the technology of the Jacquard loom. 

But however intriguing and important the technology seemed, Babbage’s Analytical Engine – due to factors financial and logistical – was never built.  It comes to us only through Ada Lovelace’s annotated translation of a French introduction to the machine – a piece of writing that established the algorithm for the computation of Bernouilli numbers, and a piece of writing that established the idea of computer programming.  Turing would later build on the work of Lovelace and Babbage, formalizing their concepts in the Universal Machine.

When Turing introduced the mathematical description of the Universal Machine in the 1936 paper “On Computable Numbers”, he set out to answer the Entscheidungsproblem, the third question left by mathematician David Hilbert.  Gödel had already answered Hilbert’s first two questions – No, mathematics was not complete, and it was not consistent.  Turing showed that mathematics was not decidable.  And that recipe to solve a particular problem, gave us an answer that begs the asking of a new set of questions.