Lab Notebook: Strange connections between baseball and Netflix

I discovered another repeating pattern for problem solving in baseball scouting and ... movie classification. In baseball, scouts and managers have to process a lot of vague information about many prospects. Nate Silver writes how Billy Beane ("a phenom in baseball management" and the main character in Moneyball, one of my favorite movies about sports) addressed the problem:

...when we have trouble categorizing something, we’ll often overlook it or misjudge it. This is one of the reasons that Beane avoids what he calls “gut-feel” decisions. (The Signal and the Noise. 2013.)

In short, Beane developed an elaborate system and a large number of explicit categories that his brain could rationally handle instead of relying on subconscious, gut-feel decisions. Using lots of  categories enabled him to pick the right player among many candidates.

Where do people experience a similar problem? In "scouting" movies on Netflix! To help users solve the problem, Netflix engineers developed a detailed content categorization system with thousands of fine distinctions, so that people can select the right movie among lots of candidates. Here's an example of their movie subjects:

Source: Alexis C. Madrigal. How Netflix Reverse Engineered Hollywood. The Atlantic. Jan 2, 2014.

A couple of learning points:
- In Scalable Innovation (Chapter 5), we talk about the concept of "Aboutness" - an element that facilitates decision making in systems. The Netflix chart above would be a great way to show how generating movie aboutness helps solve detection problems for the users.
- "Gut-feel" decisions are a poor substitute for systematic thinking about the problem. Fundamentally, they are limited by our working memory and don't scale to handle complex choice situations.

tags: aboutness, problem, solution, detection, control, pattern

Lab Notebook: Google Glass and 360-degree vision

Google Glass (or a similar connected, hands-off presentation device) should make a big difference in situations that require body-eye coordination and [related] learning of psycho-motor skills. For example, if I wear sensory clothing that talks to the device I should be able to see myself from any angle. The device might also provide an 3- or 4-D image of the desired position and help me adjust my posture closer to the ideal. Boys and girls would learn easily how to keep their backs straight and walk like models! :) Should also help the elderly to be aware of the need to exercise and improve their motor skills.
Eventually, it will become indispensable in sports, wellness, and all other applications requiring spatial dexterity, including surgical procedures.

In general, great innovations create new space and time (either real or virtual). The Google Glass (GG) project has a chance to do just that. In system terms (see Scalable Innovation, Chapter 2),

GG - Tool
sensor network - Source
reference db - Source
(wireless) communications  - Distribution
application - Control
brain - Control
encoded position info - Packaged Payload

tags: invention, space, time, problem, solution, innovation, pattern, system

Counting without numbers.

A paper in Nature Neuroscience shows that a neural network can learn how to count without any knowledge of numbers. Internally, it forms neuron firing patterns that are closer to the set theory than to arithmetic.  This research is a step toward an explanation for Approximate Number Sense (ANS), which is present in many animals, including humans (take a simple test to get a feeling for your own ANS).

08 January 2012. Nature Neuroscience (2012). doi:10.1038/nn.2996 ---
Here we show that visual numerosity emerges as a statistical property of images through unsupervised learning. We used deep networks, multilayer neural networks that contain top-down connections and learn to generate sensory data rather than to classify it^{8, 9}. Stochastic hierarchical generative models are appealing because they develop increasingly more complex distributed nonlinear representations of the sensory input across layers⁹. These features make deep networks particularly attractive for the purpose of neuro-cognitive modeling.

The ability to approximate (estimate), rather than calculate, can be critical in complex situations. It might also explain why invention of number Zero was so controversial. We have a hard time "seeing", i.e. creating an internal neural network representation, a non-existing pattern.

tags: pattern, approximation, science, network, biology

Social heartbeat.

The chart below shows relative frequency of the word "present" in Google searches in the United States (category = shopping.) Since Google is an advertising, not technology, company, its marginal financial results depend on people searching for buyable items.

tags: pattern, finance, market, social

2012-2015: lucky years to have babies in the US.

In his book Outliers, Malcolm Gladwell notes that the generation of Americans born during the Great Depression ( when the birth rate in the US was the lowest on record) did really well economically. They studied at schools and colleges that were less crowded; they got good jobs working for the Baby Boom generation families, and etc. Now, with the Great Recession entering its 5th year, we see a similar birth rate troff:

(Nov 21, 2011. BusinessWeek) - The number of births fell to an estimated 4 million last year, the fewest since 1999, according to National Center for Health Statistics data. American families -- whose finances have been hurt by high unemployment, falling home prices and low pay raises -- lack confidence to plan for “explosions in spending” required by a new child...

By 2015, women in their childbearing years may average 1.75 babies each, a 30-year low, while “ Americans still believe that two or more children is ideal,” said Bradford Wilcox, a sociology professor and director of the National Marriage Project at the University of Virginia in Charlottesville. 

Taking into account that "The annual cost of raising a child ranges between $11,880 and $13,830 for a two-parent family earning $57,600 to $99,730," parents who plan to have kids now will presumably get an immediate hit, but their children will gain a long-term advantage over other age cohorts. Further, housing prices for families with kids will soon be the lowest in a generation, which will also add to the advantage. 

From Creative to Mundane: the culture of innovation

A couple of articles about innovation caught my attention last week. One is about the rate and direction of invention during British industrial revolution in the 18th century*, another is about contemporary Shanzai culture of tech knock-offs in China**.

Both articles highlight the emergence of creative crowd made up of highly skilled craftsmen, which is necessary for accelerated tuning and implementation of breakthrough inventions. A very similar situation exists in Silicon Valley, where aboundance of technical and entrepreneurila talent makes possible the constant flow : from ideas to products. The pattern (see the diagram above) is quite general. It manifested itself in all major innovations, from cotton spinning in 18th century England, to 19th century railroads and telegraph in the US, to iPhone clones in 21st century Shanghai.

One key difference shows up, though. It mostly due to the global nature of today's technology. In England, interaction between original inventors and the creative crowd was local; in Shanghai the interaction is long-distance, i.e. things get invented in Silicon Valley and tuned to local conditions in China. As was the case in all times, technology piracy is rampant. People appropriate ideas of others and build "mashups", moving new essential skills from creative to mundane.

tags: innovation, creativity, pattern, evolution, information, distribution, source

* Meisenzahl, Ralf and Mokyr, Joel. 2011. THE RATE AND DIRECTION OF INVENTION IN THE BRITISH INDUSTRIAL REVOLUTION: INCENTIVES AND INSTITUTIONS. NBER working paper 16993. (via magrinalrevolution.com).
** Tech trend: Shanzai. @ bunniestudios.com (via Max Shtei).

The magic of the familiar.

While looking at the list of Top Grossing American films of the decade, it is hard not to notice that all of them are cinema versions of fairy tales. Moreover, it appears that movie audiences had been well primed by either a popular comic strip, a book, or previous movie based on the same characters. Technology-wise, the development of computer graphics enabled creation of magical effects that were practically impossible during the previous decade.

The Dark Night
Release date: 7/18/08
Domestic boxoffice: $533.3 million

Shrek 2
Release date: 5/19/04
Domestic boxoffice: $436.7 million

Pirates of the Caribbean: Dead Man's Chest
Release date: 7/7/06
Domestic boxoffice: $423.3 million

A graph from a 1954 article by Ernest Jawetz in Annual Review of Medicine looks remarkably similar to the Gartner Hype Cycle "discovered" in 1995. Both graphs have an enthusiasm peak, a disappointment pit, and a productivity plateau (see below).

Gartner Hype Cycle (courtesy wikipedia.org)

With Twitter we are probably still in the early stages of the cycle (see Google Timeline snapshot):

tags: innovation, cycle, diffusion, pattern, theory, book, infrastructure, niche construction, constraint

By continuities, I mean patterns that extend across time. These are not laws, like gravity or entropy; they are not even theories, like relativity or natural selection. They are simply phenomena that recur with sufficient regularity to make themselves apparent to us. Without such patterns, we’d have no basis for generalizing about human experience: we’d not know, for example, that birth rates tend to decline as economic development advances, or that empires tend to expand beyond their means, or that democracies tend not to go to war with one another. But because these patterns show up so frequently in the past, we can reasonably expect them to continue to do so in the future.

The landscape of history, by John Lewis Gaddis. p.31.

A continuity is often an indicator of an underlying infrastructure that sustains the pattern by counter-acting entropy.

tags: process, artifact, niche construction, pattern, greatest, course

Routing the signal: Nature, Nurture, and the Fed.

A remarkable similarity between the way neurons in the brain work (a nature's invention) and the way devices (a human invention) control transfer of energy, e.g. signal, in electronic devices. In the first case, signal propagation is mediated by releasing chemicals into neural environment. In the second, energy propagation is mediated by releasing electromagnetic radiation into electronic environment.

From Second Nature, by G.M.Edelman (1972 Nobel Prize in Medicine):

Each of these value systems releases a type of neurotransmitter or neuromodulator under particular circumstances. ... On receipt of salient startle signal, say a loud noise, these neurons release a neurotransmitter called noradrenaline into the surrounding space, as if from a leaky garden hose. The result can lower the threshold of synaptic responses of multiple neurons, leading to more firing as well as to changes in the synaptic strength among these neurons. p 31.

US Patent 7,587,107. Claim 1 (obfuscating terms removed):

A method of controlling energy propagation, comprising: guiding energy...; blocking the ... energy ... by electromagnetically saturating ... the guiding structure along the ... path... with electromagnetic radiation.

To control the economy, the Federal Reserve acts in a very similar way. It lowers or raises background interest rates, which either facilitates or inhibits transfer of money between economic agents.

In system terms, the Control affects certain properties of the Route(s) to change delivery of a particular Payload.

:: payload, control, distribution, problem, pattern, transfer, example

Necessity is the mother of sex.

Aphids, tiny plant-eaters, can reproduce both sexually and asexually. When there's plenty of food and other life conditions are good as well, they

"increase exponentially via asexual reproductive strategies to take full advantage of the rich supply resources.
When food sources have been depleted, the climate becomes hostile, or individual survival is jeopardized by some other adverse change in living conditions, these organisms switch to sexual forms of reproduction. Sexual reproduction ensures a mixing of the gene pool of the species.

Does technological innovation follow the same set of strategies? When funding is good every startup seems to  develop Web 1.0 or 2.0 or whatever the latest buzzword-compliant application is. When funding gets scarce or changes in the environment come fast and furious, people (even the military) tend to become more willing to mix things up, broker, rather than kill, ideas, and transfer technologies from different fields of use. In some ways, layoffs in hi-tech can be considered as a way to reshuffle knowledge and skills.

In January, 2009, New Scientist published an article about the current debate in evolutionary biology on Darwin's "tree of life". The emerging view is that biological organisms do not form a neat genealogical tree, but rather a network that is ruled by Horizontal Gene Transfer (HGT).

Darwin assumed that descent was exclusively "vertical", with organisms passing traits down to their offspring. But what if species also routinely swapped genetic material with other species, or hybridised with them? Then that neat branching pattern would quickly degenerate into an impenetrable thicket of interrelatedness, with species being closely related in some respects but not others. ...

As more and more genes were sequenced, it became clear that the patterns of relatedness could only be explained if bacteria and archaea were routinely swapping genetic material with other species - often across huge taxonomic distances - in a process called horizontal gene transfer (HGT).

Technology, being closer in complexity to microbes rather than humans, seems to develop by transfer of ideas between different domains. The broker innovation model described in detail in "How Breakthroughs Happen" shows multiple example of solutions transfer from high-tech to low-tech industries. Furthermore, the Theory of Inventive Problem Solving (Russian acronym TRIZ) is built around the concept of repeating invention patterns that can be extracted from patented solutions. Also in this mold is template-based creativity approach proposed in Creativity In Product Innovation.

Though the mainstream innovation literature tends to focus on the personality of the inventor, real invention/innovation research work should be directed toward discovery of solution patterns and their proliferation throughout the technology noosphere.

First version is always a hack.

New Scientist has a 50-year retrospective about evolution of the spacesuit.

The suits were adapted from US Navy pressure suits for high-altitude flights and not designed to be worn on spacewalks.

We can see how a design, borrowed from a different application domain, over time is adapted to new tasks specific to space exploration. Now, compare this development to the evolution of the smart phone. The device is still being sold as a phone, i.e. voice communication appliance, while in the future it will evolve into a vehicle for application and data space exploration in the world of networks of networks.