Our Man in Nirvana

Rate of Innovation Part II

As a child I was a keen chess player but I have not played regularly for over 20 years.  That has changed recently with technology giving me two new ways to play and practice.  First, I have bought a chess computer game on my Sidekick II.  I can now play a game in 5 minutes against a computer during idle times such as waiting for MUNI.  Second, Auren Hoffman invited me to play online at Red Hot Pawn where you can play multiple asynchronous games against other people over the web.  I am still playing my first game with Auren which started on February 19th.  I must admit I am finding the lack of rhythm a challenge with a game spread out over several weeks but it's good to be playing a human being rather than a computer

Speaking of human beings vs computers, as I have been laid up in bed for a couple of days trying to shake off a virus, so in between sleeping and drinking lots of water I took the opportunity to watch the DVD Game Over about Gary Kasparov's chess match against IBM's Deep Blue in 1997.  The premise of the documentary is intriguing - IBM cheated to beat Kasparov to bolster their flagging image - but the execution and evidence presented is circumstantial and hearsay at best.  Putting aside the premise and just examining the trends - Kasparov beats computer in 1996, Kasparov loses to computer in 1997, Kasparov draws to computer in 2003 - the seemingly slower rate (backwards?) of innovation is in marked contrast to that shown in the DARPA Grand Challenge.  Why is this?

The reason that first comes to mind is that the tasks differ in their complexity.  My gut instinct is that the task of getting a car to autonomously drive from A to B on a known course is less complex than defeating the world's greatest chess player with more playing permutations than there are atoms in the universe.  However, upon further reflection it's not clear that the number of known permutations is the best measure of complexity.  The number of chess permutations is enormous (10 to the power of 120) but there is no element of randomness, unlike the DARPA Grand Challenge where the cars have to deal with other cars, weather conditions, changing gradients, etc.

I shall noodle on this some more, but it's a cautionary tale on extrapolating from just a few data points to a wider conclusion.

Rate of Innovation

This past week I attended the 4th Annual MediaX Conference at Stanford.  MediaX is the organization that strives to connect industry with the interdisciplinary research on interactive media going on at Stanford in 18 different schools and research centers.

There were many interesting ideas on show during the day and a half event, but  the presentation that stood out was by Computer Science Professor Sebastian Thrun when he told the story of Stanford's participation and victory in the 2005 DARPA Grand Challenge.  In 2004, DARPA (ie the US military) had set a challenge for an autonomously driven car (ie no human driver either within the car or by remote control) to complete a 100 mile course across the dessert.  The participants would be given GPS coordinates 2 hours before the race to input into their systems after which they would have no direct control over the vehicle.

In 2004, the best performing car completed 5% of the course.  In 2005, 5 teams completed the entire course.  So in one year performance was improved by 20X (not the boring 2x we are used to under Moore's Law) and this was achieved by a group of undergraduates!  That's a sobering thought when we are trying to extrapolate from now to the future.  Even more impressively, Professor Thrun has set his team the challenge of successfully having an autonomously driven car from downtown San Francisco to downtown Los Angeles by October 7th, 2007.

Nuclear Power, Climate Change, and the Next 10,000 Years

At my wife Rachel's suggestion, on Friday night we attended the discussion titled "Nuclear Power, Climate Change, and the Next 10,000 Years" at the atmospheric Herbst Theater between Peter Schwartz (co-founder of the Global Business Network and the father of scenario planning which he did at Shell) and Ralph Cavanagh (co-director of the Energy Program at the National Resources Defense Council).  The event was hosted by the Long Now Foundation  who are striving to provide a counterpoint to today's "faster/cheaper" mind set and promote "slower/better" thinking.  It was my first exposure to this organization, but I know it shall not be the last given the topics and quality of the participants.

The format was not a classic debate structure. Instead it required each of the speakers to first listen to their opponent state his position, then ask questions of the opponent to refine his position, and then finally restate his opponent's position to the opponent's satisfaction.  The floor was then open to questions from the audience and the moderator Stewart Brand.

Peter Schwartz went first and articulated his position as follows:

1. Continued emissions of CO2 and the inherently unstable nature of the earth's temperature will lead to the risk of civilization coming to an end within the foreseeable future.

2. Continued demand for energy, with the emergence of another 1 billion people coming out of poverty led by China and India, and the decline in oil production will lead to tremendous pressure on other forms of energy (nuclear, wind, solar, biomass)

3. Only nuclear energy has the potential to i) avoid dramatic climate change ii) and fill the demand-supply gap (requires "only" doubling current world wide nuclear capacity of 400 reactors, compared to >70x increase for any other technology) and iii) advances in technology can address the historic safety and waste management issues.

4.  Therefore the US government should show leadership and put nuclear power back on the table as the likeliest energy solution while avoiding cataclysmic climate change

Ralph Cavanagh's position was

1.  Ditto

2.  Ditto

3.  The failure of nuclear power as a viable alternative to coal or oil is clear and unlikely to change - no new nuclear power plants ordered in the US since 1973, lack of cost competitiveness, regulatory tug-of-war over Nevada's nuclear waste dump has dragged on for decades, $150 billion dollars of development and the problem of waste has not been solved, the risk of empowering rogue nuclear states, etc

4.  Therefore the appropriate role for government is not to anoint nuclear a priori as the best solution, but provide a regulatory framework that includes emissions as a true cost and let the competing technologies (coal sequestration, nuclear, wind, solar, biomass) fight it out in the market place.

For those wanting to get the full details of the debate, it will eventually be available for download here.  In the meantime, it made me thankful for my intellectually curious wife, it reinforced our desire to get a Prius (we've made the decision, just waiting for our preferred colour to be available) and learn more about the potential alternative energy technologies, it made me feel good about the Mobius investment in IonAmerica (a company trying to address the commercial power generation market through solid oxide fuel cell technology), and it encouraged us to add the National Resources Defense Council to our potential list of donations.

Not bad for a Friday night.

Googlearchy

One of the issues I wrestle with  when surveying the Web 2.0 investing landscape, is whether a new company has the ability to challenge incumbent established companies.  The term Web 2.0 is not fully defined, but for me it refers to technical (AJAX, RSS) changes and consumer behavior (broadband adoption, user generated content) changes that have allowed successful companies to be built for much less capital.  For example, Ofoto is Web 1.0, Flickr is Web 2.0. 

However, while I believe that most of the economic trends are all moving in the right direction to build and launch a Web 2.0 business (decreasing bandwidth, processing and storage costs) I do worry about Googlearchy.  This is the concept of Google's PageRank algorithm reinforcing traffic patterns whereby the "rich get richer" and imposing a structural disadvantage for new websites.  Given PageRank is driven by the number of incoming links and the importance of those links, this concern seems to be confirmed both intuitively and mathematically, but until recently this has not been empirically tested.   The first test of this "common wisdom" has been made by Santo Fortunato and his colleagues from Indiana University in a paper called "The egalitarian effect of search engines"  published in arXiv.

Fortunato and his colleagues built two theoretical models - one where all searches are done through search engines, one where all searching is done through surfing - and compared the results against empirical data based on 28,164 randomly chosen websites.  The surprising result was that the empirical data did not lie between the two theoretical extremes.  Fortunato's conclusion from this observation is that search engines, rather than giving structural advantage to incumbent popular sites, actually level the playing the field.  The paper has since come under criticism for both the robustness of the data set (ie "different data sets will not produce the same result") and the causality ("even if the new data sets produce the same results, the result is not due to search engines").  The criticism is understandable given the paper questions "common wisdom" so I look forward to watching how the debate progresses given my vested interest as a VC looking to invest in Web 2.0 companies and overcome the "Googlearchy".