262ventures.com

Two of my favorite books are James Surowiecki’s The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations and Charles Mackay’s Extraordinary Popular Delusions and the Madness of Crowds (fun fact: MacKay’s book, when originally published in 1841, was Extraordinary Popular Delusions and the Madness of Krauts. Apparently, this title fell out of favor at some point). Surowiecki’s work argues that the collective wisdom of groups is superior to the knowledge of any individual in that group. Mackay tome instead chronicles the frequent instances in history in which the crowd has practiced a sort of collective insanity (frequently in the form of some sort of bubble). So, which is it: are crowds wise or crazy?

There are many people smarter than I that are thinking about this apparent paradox (including Surowiecki himself), so I won’t claim to add anything substantive to the debate. However, I think there are some really interesting analogies to be made to Ben and my own chosen field of chemical engineering. According to Surowiecki, four elements are required to form a wise crowd: diversity of opinion, independence of opinion, decentralization (people can draw on local knowledge), and aggregation (some mechanism to collect this knowledge). From an engineering perspective, those four qualities are the same four qualities of an ideal gas - basically, a dilute mixture of gas molecules moving about via random Brownian motion. Although the molecules of an ideal gas move about randomly (independence and diversity of opinion), collectively, they move so as to maximize the entropy of the system (aggregation). And, from information theory, entropy=information, so that in maximizing entropy, the gas molecules (or, in our case, investors), are maximizing the information content of the system. Although no individual gas molecule knows anything about the entropy, collectively they maximize it. Sounds like the efficient market hypothesis, no?

So, if the wise crowd is a container of ideal gas particles, when does the crowd go mad? In keeping with our analogy, the crowd goes mad when someone turns on a fan. Now, all our gas molecules are moving in one direction, and brownian motion is overwhelmed by convection. Maybe the fan is the media, a new investment product, a new technology, or something. But when we turn on the fan, all bets are off.

End nerd talk.

So Clint Eastwood gets canned from the California State Parks Board, presumably for his opposition to the creation of a new toll road that would run through San Onofre State Park.  The governator argues that the toll road would do much to relieve the congestion in the OC.  He’s probably right.  But here in Southern California where it takes an hour to get twenty miles sometimes (according to Bill Simmons), additional bandwidth may not be the answer.  Indeed some predictions have indicated the population of Los Angeles may double over the next few decades.  I doubt that a six-lane tollway will do much to accomodate the approximately new 3.5 million people from LA on their way down to Huntington Beach.

Los Angeles and the surrounding counties may prove to be an experimental playground for urban planners and landscape architects over the next fifty years.  The proposals ring of bold and forward thinking regarding how spaces should interact and connect.  Combine these developments, which may shorten average car rides, with novel technologies for transport management, such as an integrated system of sensors and controls governing traffic and we may get the congestion relief we need.

I’m all in favor of improved infrastructure and greater efficiencies.  But focusing on what most definitely is a short term solution is myopic.  Keeping San Onofre a little more pristine will be just a side benefit to a well thought out long term plan.

I have been itching to read Dan Ariely’s Predictably Irrational for weeks now.  I haven’t ordered it on Amazon because my wife has put me on book-buying hiatus, suggesting that I should support my local public library (that’s another story).  It sounds like great fun and an interesting look inside our own CPUs.

So with the book in mind I came across this article earlier today.  Apparently UT researchers have demonstrated that the more attractive a woman is communally regarded as, the more selective she is about that attributes of her mate (which include earning power, physical pulchritude, familial skills and loyalty).

This is in stark contrast compared to men.  Erik reminded me that Professor Ariely was one of the co-authors of What Makes You Click? Mate Preferences and Matching Outcomes in Online Dating.  In the paper (which discusses women and men through the veil of online dating), their study isolates a particular parameter for the male criteria as a potential mate.  It’s pretty heady stuff: apparently dudes prefer hot women.

I wanted to follow up on the previous post regarding pandora.com. Pandora represents a bit of an anachronism in today’s world of artificial intelligence, a relic from our linear regression past. Our what now? Well, back in the good old days (say, the 1990’s), people did research where they measured some outputs (in Pandora’s case, how much people like a given song), and then correlated them to some inputs (for Pandora, the qualities that make up the song). The statistical method one uses to do this correlation is linear regression - we assume that if we like a little of something, we’ll like more of that something even more. This is what Pandora does - it learns that you like electric guitar riffs, or “unintelligible lyrical style” (yes, this is an actual classification in use by Pandora). It then suggests songs that also have these qualities. It’s recommendations are based on the painstaking work of over 50 full-time employees, whose sole job is to listen to music and categorize it on dozens of different characteristics.

These days, recommendation algorithms are much more likely to follow the Amazon algorithm - call it our networked present. Amazon generates your recommendations based on what similar people have purchased. For example, I recently purchased Nudge: Improving Decisions about Health, Wealth, and Happiness by Richard Thaler (which looks excellent by the way). Because of this, Amazon thinks I might also like Sway: The Irresistible Pull of Irrational Behavior by Ori Brafman. This seems reasonable, I suppose: if I am interested in the behavioral finance of nudging, I would probably be interested in the behavioral finance of swaying as well. The networked present is so popular because it is so easy. Amazon doesn’t need a full time staff of reviewers classifying books as “behavioral finance”, or “one word titles followed by a colon”. The purchase data is just sort of…there. No need to go out and get it.

For all their simplicity, networked algorithms have some significant weaknesses. They tend to promote closed systems - if I write a new book entitled “Shove: The Economics of Smackdowns”, someone needs to buy both my book and Nudge before the network will make the match. Networked algorithms also have a natural tendency towards the popular - since a lot of people have by definition bought the popular item, it will show up frequently in the recommendation algorithm, while the less popular option will go unnoticed.

What is a person to do? Well, the best sites use a little bit of both algorithm types. For example, Netflix has devoted considerable resources to their recommendation algorithm, which uses both network effects as well as linear effects like the genre of the movie and the main actor(s). Amazon also allows you to rate and categorize products, to improve your recommendation. In sum, it pays to practice algorithm inclusivism.

Erik, Jeff (his brother and one of the front men to the Plastic Constellations), and I emailed about about the Pandora  music radio service starting a few weeks ago and I’m including it here:

3/7/2008 (Erik)

Have you guys ever checked out pandora.com?  It is internet radio,
but they create a radio station for you based on an artist list that
you provide.  It is part of the “Music Genome Project”, which
attempts to extract information from music samples, and search for
similar artists.  Pretty fascinating concept.

Anyways, I ran an experiment, which I found interesting.  First, I
created a “TPC” (The Plastic Constellations)  station.  The first song was “Beats Like You Stole
Something”.  So far, so good.  pandora tells you what key attributes
they have for a given song.  For Beats, it was:

electric rock instrumentation
punk influences
a subtle use of vocal harmony
mild rhythmic syncopation
repetitive melodic phrasing
extensive vamping
a vocal-centric aesthetic
minor key tonality

Pretty good so far, right?  However, all the other music they gave me
was crap.  ”Heamiser” and “Fireball Ministry” - it seemed to be a lot
of 80’s ballad type stuff.  So, I added more artists.  I tried to add
the following 3:

Thunderbirds are Now!
The Hold Steady
Sonic Youth

Don’t know why I picked those, they just came to my head.  This sort
of worked, but the music was a lot of Nirvana, Smashing Pumpkins,
etc.  Things I enjoy, but not really what I had in mind.  My solution
was to remove Sonic Youth - I figured they were too famous, and
messing up the comps.  When I did that, I actually got a pretty good
station.  The last few artists have been:

The Bravery
Mon Frere
Mendetz
Ima Robot
The Hold Steady
Cold War Kids

Now, this is where my indie rock ignorance gets in the way.  Are
these good comps?  What bands should I try adding to improve the
fit?  I find this whole thing fascinating.  Any ideas?

3/7/2008 (Erik)

I have another algorithmic idea that I want to work on, but I doubt
there is a lot of commercial potential - genre adjusted tomatometer
scores.  See, I love Rotten Tomatoes, but the actual % fresh is
pretty useless I find, because critics like certain genres, and
dislike others.  If a movie is a documentary, it can get 90% approval
for just showing up.  Similarly, any CGI movie gets 90+% no matter
how good or bad.  By contrast, comedies (and if I may make a claim of
subtle racism, especially comedies starring African Americans) get
the shaft unless they are amazing.  Basically, we just need to
compute the averages of each genre, and give a scaled tomato-meter
score.

3/20/2008 (Erik)

Sorry, I promise this is the last Pandora email (unless I find more
cool stuff).  Anyways, I have been listening to the TPC station for a
few weeks now.  When I hear a song I like, I give it a thumbs up.
When I hear a song I don’t like, I give it a thumbs down.  At this
point, the radio station is pretty good - I would say I enjoy 80% of
what comes on now.

So, I listened to 15 songs, and for each one I copied down the
characteristics Pandora said it had.  Here are the characteristics
that appeared most often:
a subtle use of vocal harmony
a vocal-centric aesthetic
electric guitar riffs
electric rock instrumentation
mild rhythmic syncopation
punk influences
repetitive melodic phrasing
varying tempo and time signatures

At this point, the thing I find very interesting is how my radio
station now tells me _what I like_.  Pandora is basically an AI
algorithm, and now that it works, it can tell me explicitly what I
like in a song.  I find it interesting, because I agree with all of
the characteristics above, but I would not have been able to make
that list for you before this.

Could Pandora be used to determine what the general public would like
in an album, and is it already being used for that?  Interesting
question, I think.

Nature Magazine’s latest issue (Mar 20 2008)  has a special section on the water situation including a nice article about novel technologies to address potential crises.

But equally interesting is an article on a variation of the prisoner’s dilemma game, often taught in courses on game theory.  The results are fascinating to say the least.  The authors (Dreber et al.) seem to have found that in a social milieu which involves communication in one of three ways (cooperation, defection, and harsh punishment), the players that used the harsh punishment received no additional economic benefit.  Instead, the non-punishers tended to come out ahead on the scoreboard.  This seems to hvave potential significance regarding how teams function.  I wonder if this can be applied to sports teams?  Is the value of a Steve Nash or Chris Paul assist higher than anticipated?

Who doesn’t like clever thinking?  There are a bunch of ideas which, upon retrospective analysis, you kick yourself for not having thought of first.  Recently an osmosis based power generator has been commissioned in Norway by the utility company Statkraft.  Osmosis based power generation is essentially to flip the process of desalination via reverse osmosis (RO) membrane filtration.

In RO membrane desalination, salt water is pushed through a membrane which prevents the passage of salt.  On one side of the membrane you have clarified water and the other brinier water.  The system does not prefer this non-equilibrium state and accordingly the clear water and salt water want to mix back together (this is what is known as a difference in chemical potential on the two sides of the membrane and gives rise to what engineers call osmostic pressure).  Because water can flow back through the membrane to remix with the salt solution, energy in the form of pressure must be input to the system to prevent water from flowing back.

In osmotic energy generation, clean water is brought in through the membrane to the salty side.  Water easily flows through the membrane to mix with the salty side and generate work.  Simple right?  Clean, no emissions, and abundant salt water and …

Oh wait… clean, clarified, fresh water is already a challenge to obtain for billions on this earth.  You can’t get it everywhere.  And if sea level rise, low lying fresh water will become brackish.

The point is that you can’t get something for nothing.  This is the lesson that we took from grad schoool in Jeff Tester’s Introduction to Thermodynamics 10.40 class at MIT.  In real-world application, there is no such thing as a reversible process.  In RO desalination, we fight entropy using work.  In osmotic power generation we leverage entropy to create work.  It sounds like a perfect strategy if you can create one site to do both the forward and reverse operations (generate energy or clean water when one or the other is required) But between those two steps, we’ll always lose to real-world losses.

I’m really excited for two potential collaborations that are brewing up now.  One involves potentially working with a professor at MIT (my little sister) and a professor at Stanford (buddy from Stanford).  This is the sort of thing I dreamed about when I was in college - working with friends and family.  As we all started working/going to graduate school/doing research and saw our fields start to intertwine, the possibility began to solidify.  So we’re giving it a shot and we’ll see if this project can come together.  The project should involve energy distribution.  More on that later.

The other collaboration is coming together with a couple buddies from graduate school.  We’re putting some tools together to address some health related problems and I think this could have a strong potential impact in improving the distribution of information and health care.

Many spin-off companies that arise out of academia raise funds through government/public grants such as small business innovation grants (SBIRs).  These SBIRs serve to provide funding for entrepreneurs to help get their technology out of the lab and into the proverbial garage.  However, the bureaucratic hoops that one must jump through, followed by diversion from developing the core product, can wind up taking a company on the fast track to a meandering eternity of ‘SBIR house purgatory’.

The alternative is to get VCs on board right away, raise a few million, and focus on developing the core product.  Sure you might wind up with only 20% of the company left to split among four founders but hey, you’ve got an accelerated product build out and a 3yr exit strategy right.

The right mix of funding, as well as the total amount of time-resolved funding, is critical to novice entrepreneurs.  Few startups share this in common and is probably harder to get right during startup cocktail hour.

Scientific debate should be exactly that - a debate. Correspondingly there should be sufficient room in the discussion for both proponents and opponents of the global warming thesis. Indeed many of the major points offered by the counterpoint seem valid - for example why risk economic prosperity for reasons unproven in real scenarios?

Underlying the scientific process is belief. You have a hypothesis that you believe in and then, as a scientist, you attempt to observe the event in a controlled manner, in the absence of interfering effects. Theoretically, if you can isolate the dominant factors for an observation in a logical manner, your belief is validated and accepted by the scientific community.

The difficulty in analyzing the phenomenon of global warming in the context of human  generated carbon dioxide is that the Earth’s climate is indeed complex. The data is often fuzzy because the timescale for effects to manifest typically outlast generations. Thus changes in temperature from year to year may lie within the noise. Climate change skeptics have hopped on this, in addition to the notion that predictive climate model prediction have not adequately reproduced reality, to assert that climate change is merely a blip in the road. This is what they believe and what could be true. I disagree, but it may be true.

However we must be careful of what the noise tells us. As scientists, we should aggregate what we know from controlled experiments, extend them to their logical conclusion, and assign probabilities for those events to occur. In the case of global warming the probability is reasonable that there maybe cataclysmic consequences for the environment - not just global warming but changes in oceanic conditions.  We should prepare for those consequences. Just because earthquakes occur _unpredictably_ does not mean that we don’t store extra water and food here in southern California. And sure the current physics in our models have not told us everything about climate change. But can’t the same can be said for our econometric models which now all corporations, governments, and populations treat as beacons of light during recessions?

After all scientists, as all humans, make mistakes of belief all the time. Galileo was the right voice dissenting against the beliefs of the geocentrists. But so was Einstein when he argued against the quantum mechanical model.  Beliefs have been proven and disproven through experiment and logic.  But while we could dawdle and enjoy the ivory towers in those situations, it seems we must move to counteract global warming.  While the notion that the world was the center of the universe was important for belief, its impact on the world was slight.. not so with global warming.

[disclaimer: I (Ben) believe that human generated carbon dioxide poses a huge risk for our world and environment: it has been in demonstrated in the lab to be a heat trapping gas and a factor in the acidification of the oceans (anyone getting milliQ DI water will attest to that, noting a pH of 4.7 after a few minutes in air).