“It is possible, in life, for everyone in the world to be wrong.”

…is what Nicholas Negroponte used to tell his son.  On October 13, I got to hear Nicholas speak for the MIT Club of Northern California about his latest project called One Laptop Per Child. As director of the MIT Media Lab, Nicholas founded for-profit companies such as Wired Magazine and went against the advice of everyone. He made OLPC into a non-profit so that he could achieve clarity of purpose that would enable partnerships with organizations and governments around the world that a for-profit could not achieve:

To create educational opportunities for the world’s poorest children by providing each child with a rugged, low-cost, low-power, connected laptop with content and software designed for collaborative, joyful, self-empowered learning.

By doing so, he found that they could attract the best industry talent to OLPC that they could never afford to pay for and OLPC is at 23 full-time-equivalents (FTEs). OLPC is launching these laptops to children in 31 countries around the world (like Rwanda, Mongolia, Haiti, Peru, Uraguay, India, …) with software and lessons in local languages. Prodction is at 100K laptops per month, and their eventual target market is to reach 500 million children worldwide who lack access to computers.

These computers have a mesh Wi-Fi network interface which is connected to a satellite hookup to the Internet where available — so in addition to Internet access, the laptops can automatically check-in to a central server in each country to track their status. If my memory serves me correct in some countries like Peru and Rwanda he reported 100% of the laptops check-in, in others they have not seen a large fraction of the laptops.

The principles driving their design and deployment of these laptops are:

1. child ownership — low cost laptops designed with a child in mind
2. low ages (6-14) — view children as a market
3. saturation — access for everyone not just a few computers in each school
4. connection — to the Internet
5. free and open-source software

The idea for this started out when Negroponte and his colleagues at MIT recognized that computer programming was the closest we get to thinking about thinking, and when computers are out in the hands of children in developing countries they can learn something not possible with any other means. It brings the learning experience alive for children and teachers even in traditional classrooms.

Going forward, scaling is their biggest organizational challenge. Technically they need to increase the level of integration for the hundreds of components in their laptop to bring down cost. They do not see themselves as being the primary provider of content (lessons, applications) — the analogy he used was that Gutenberg didn’t write the books. He says they don’t compete with Microsoft/Intel in the same way that the UN World Food Program doesn’t compete with McDonalds.

This holiday season OLPC is introducing a Give-One-Get-One program to allow people to buy an OLPC laptop that pays for (donates) a laptop to a child in a developing country.  For their evential target of 500 million children, they need $50 billion (500 x $100 million) to saturate the world with laptops. I think coming up with that amount of cash will eventually become the limit to OLPC’s scaling unless they find some other source of funding through government or private foundations.

Nicholas started off his talk by telling a story on how he got funding for starting the MIT Media Lab that he said was never told before. The MIT president at the time was nearing the end of his term and decided he wanted to do research instead of moving up to Chairman of the MIT board. So Nicholas decided this was an opportunity to start a lab, but needed some funding to do so — where to start? He called up the chairman of NEC saying that others in France would invest provided the chairman invests. Once the chairman decided to invest, he said the same trick on his other investors and so he ended up with the funding he needed.

Like Nicholas, many great people I have admired have at one time or another said something like what Feynman did — “What do you care what other people think?”  Investors like Buffet, Soros, and Trump also say something similar about being contratian. Here is a link to a video from 1984 where Negroponte describes his visions for the future.

update on 5/2, see “Einstein the nobody” and  “The World As I See It”

Advertisement

Phrases, natural language processing, and artificial intelligence

Michael Arrington’s long interview with Barney Pell and the Microsoft product manager for search after announcing the Powerset acquisition gives hints as to the Powerset’s capabilities that MSFT search needs most vis-a-vis what Google already has incorporated into their search. A lot of it seems to fall into the “phrase” statistics category instead of “linguistic” understanding category — see below but also see this discussion about Powerset and other players like Hakia in semantic search. Given the performance of today’s mainstream search engines from Google/MSFT/Yahoo, I’m beginning to think that until the day we have artificial intelligence, phrases (bigrams, trigrams, etc) and related statistical techniques will remain the silver bullet for general purpose web search.

The Powerset/MSFT folks have a BHAG (big-hairy-audacious-goal) and it’s hard to fault them for that. The technical challenge that Powerset is running up against is the same sort of wall (i.e. steep learning curve) that shows up in modern computer vision and image recognition applied to things like image-search engines or face-recognition in photo collections. The point here is that the technology works in well-defined problem domains or with sufficient training/hints, and it is also massively parallelizable on today’s low-cost server farms for specific tasks — it either fails miserably or has degraded/unreliable performance when the software/code is not programmed to deal with increased scene or corpus complexity, i.e. combinatorial explosion of inputs that arises with diverse features. This happens when the scene or corpus is noisy and when scene/corpus features are ambiguous or new — exactly the situations where human beings still excel in making sense of the world through judgments, and why humans still write software. For example, see these papers from Stanford computer science class projects (1 and 2) that show how unreliable the face-recognition step can be for auto-tagging Facebook photo collections and in-turn how training can be used to achieve “100% Accuracy in Automatic Face Recognition.”

First of Powerset’s capabilities that MSFT wants is word-sense disambiguation — both inside user queries and in the corpus text — to produce more precise search results. Thhe example he gives is “framed.” Two attached papers listed below from the 1990s talk about two different approaches using statistical collocations (Yarowsky) and latent semantic indexing (Schuetze)… this is something MSFT may not be doing yet today, but Google is doing using perhaps Bayesian word clusters and other techniques.

RN: I think everything that Barney said was right on. I think you see search engines including Live Search and also Google and Yahoo are starting to do more work on this matching not exactly what the user entered but it is usually limited to very simple things. So now all of us do some expansion of abbreviations or expansion of acronyms. If you type “NYC” in a search engine these days, in the last couple years, it understands that it means the same thing as New York. These are very very simple rules based things, and no one understands that bark has one meaning if it about a tree and a different meaning if it is about a dog. Or an example that someone gave the other day was the question of “was so and so framed.” And framed could mean a framed picture or it could mean set-up for criminal activity that did not occur, and so on. And you have to actually understand something of it is a person’s name then it applies to one sense of the word framed if it is not then it doesn’t. So one of the things that Powerset brings that is unique is the ability to apply their search technology to the query to the user’s search in ways that are beyond just the simple pluralization or adding an “-ing” is that Powerset also looks at the document, it looks at the words that are on a web page and this is actually very important. If you look at just the users query, what you have available to you to figure what they are talking about are three words four words five words, maybe even less. That can give you certain hints. If you look at a web page that has hundreds or thousands of words on it you have a lot more information you can use if you understand it linguistically to tell what its about, what kind of quieries it should match and what kind of quieries it shouldn’t match. And Powerset is fairly unique in applying this technology in the index on a fairly large scale already and with Microsoft’s investment and long term commitment we can scale this out even further, an apply it even more of the web, not just the wikipedia content they have thus far.

Second is the potential to support complex but realistic queries in the future — they suggest that this can replace vertical search engines which seems a bit far out.

RN: I think what Peter Norwig is saying has some degree of accuracy and that he is also ignoring some things. So, just for normal queries, queries that are not phrased as questions, there is a lot of linguistic structure. If someone types in a query that is “2 bedroom apartments, under 1000 dollars, within a mile of Portero Hill.” That query is loaded with linguistic content. And that’s a realistic query. That is the type of thing that customers actually want to find on the web. Today there is a sort of helplessness, where customers know that certain queries are too complicated, and they wont even issure them to a search engine. They will go to some deep vertical search engine where they can enter different data into different boxes.

The “2 bedroom apartments, under 1000 dollars, within a mile of Portero Hill” example is interesting, since it requires logic and can’t be solved just using related keywords. Even with access to infinite compute cycles I don’t expect any general purpose search engine applied to the web corpus can do this query reliably today, including Powerset. For verticals like real-estate or travel, we will still need things like Craig’s list or Orbitz/Mobissimo that list classified ads and or build on top of user submissions, metadata, and targeted crawling.

Compared to phrases, linguistic analysis helps provide a better answer to the question of whether a given web page matches the user’s intended query or not. It certainly has the potential to improve recall through more accurate query interpretation and retrieval of web pages from the search engine’s index. By itself linguistic analysis does not improve the precision of search results (by re-ranking) in the way that hyper-link analysis did in Google’s PageRank and other web search engines.

The third use case is query expansion to related keywords using an automatically-generated thesaurus — which I thought Google is already doing.

nce; that is not really an area that is that interesting. But some of these more complex queries really are. For example, shrub vs. tree. If I do a search for decorative shrubs for my yard, and the ideal web page has small decorative trees for my garden, it really should have matched that page and brought it up as a good result. But today Google won’t do it, Yahoo won’t do it, and Live won’t do it. So even in these normal queries there is a lot of value in the linguistics.

I didn’t understand this “decorative shrubs” example. When I try it out on Google I find a result that seems relevant to me. Furthermore using the “~” option in front of any word, Google allows you to search using “related” keywords. If you try “decorative ~shrubs” you get results containing “decorative plants” as well.

Fourth, it seems 5% of queries are linguistic queries today (versus phrase queries), and they expect that fraction to expand as MSFT’s search engine capabilities mature to incorporate linguistic understanding of the query — mostly related to facts or solving problems.

BP: Let me answer the question of does anybody actually search this way. The answer is yes, people do this. It isn’t the most common mode, but we do see that probably 5% of queries are natural language queries. These are not all queries that are phrased in complete sentences, but they are queries where the customer has issued something that has some sort of linguistic structure. Almost any query with a preposition: X and Y, A near B, attribute A of Y, etc. Those things are loaded with linguistic structure…. RN: I have a list of some natural language queries in front of me. Can we just show you some queries that our customers have actually sent to us and are random examples. The first person to see the dark side of the moon. How to get a credit card in Malaysia. Enabling system restore in group policy on domain controller. Timeline of Nvidia. How to measure for draperies. What is the difference between Mrs. and women’s sizes? Does my baby have acid reflux? I could just go on and on and I. These fit in the category that we’ve labeled that match about five percent of queries and they’re really just cases where the customer can’t think of a simpler way to express it.

Finally, the people use case seems to be one of Powerset’s most visible strengths…

BP: We return answers. We actually synthesize, so if you were to say, “What did Tom Cruise star in,” you actually get not just the movies, but the cover art for the different movies. It synthesizes multiple pieces of information to give you a whole different kind of presentation. Or, if you were just to say, “Bill Gates” you’d be given an automatically generated profile of Bill Gates, pulled across many, many articles. It’s no longer just about 10 links, although we can certainly do more relevant job (and will) of the blue links, and a better job of presenting those links.

A friend of mine tried getting answers via Google, and got quick (< 5 minutes for all the searches) and relatively accurate answers without much hassle using phrases:

• “far side moon” (see section on Exploration)
• “get credit card Malaysia”
• “enable system restore group policy”
• “Nvidia history”
• “drapery measurements”
• “misses vs womens”
• “baby acid reflux symptoms”

Those are examples that illustrate why phrases and statistical approaches are not only “good enough” for searching the web corpus, and but they have a very high “signal-to-noise” compared to all other machine search techniques that became available to us in the last 20 years.

The root issue seems to me to be that human symbolic and image comprehension is still well beyond today’s programmable machine capabilities — which has nothing to do with compute power or specific interface/modality (language or vision) — it has much more to do with human intelligence itself that resides inside the cerebral cortex and other parts of the central nervous system. I’d be willing to bet that to achieve their BHAG and make the real quantum leaps from today’s search technology that they advertise, Powerset + MSFT will need the equivalent of artificial intelligence, not just natural language processing (or better image processing in the case of scenes).

Powerset is not the only semantic search engine. Besides Hakia, there is Cognition Technologies — a company founded by a UCLA professor — that also has their own wikipedia search http://WIKIPEDIA.cognition.com. This white paper compares their results to Powerset, query by query, showing how Cognition’s precision is higher and most of the time recall is much lower (more relevant vs unrelated hits) — of course this is a test designed by Cognition, and therefore a display of their own strengths. In one example the show Cognition returning exactly one result and Powerset returning 259. Also see Top-Down and Bottom-Up Semantics by Cognition’s CTO. Cognition’s other white paper compares them query-by-query to Google, Yahoo, Verity, and Autonomy — see section VI. In these tests, only Google has an observable “statistical” boost but doesn’t quite work as well as Cognition — more precision and higher relevant recall. In summary, Cognition achieves these results using the semantic processing and natural language parsing technology described here,

Cognition’s Semantic NLP Understands:

• Word stems – the roots of words;
• Words/Phrases – with individual meanings of ambiguous words and phrases listed out;
• The morphological properties of each word/phrase, e.g., what type of plural does it take, what type of past tense, how does it combine with affixes like “re” and “ation”;
• How to disambiguate word senses – This allows Cognition’s technology to pick the correct word meaning of ambiguous words in context;
• The synonym relations between word meanings;
• The ontological relations between word meanings; one can think of this as a hierarchical grouping of meanings or a gigantic “family tree of English” with mothers, daughters, and cousins;
• The syntactic and semantic properties of words. This is particularly useful with verbs, for example. Cognition encodes the types of objects different verb meanings can occur with.

Also see Q-Go, a European NLP technology company for customer service applications that partners with MSFT,

The intelligent, linguistic technology-based software that Q-go develops makes sure that customers receive a direct, immediate answer to all of their questions from a company’s website. Online, in their own language, and with at least the same comprehensiveness and quality as the answers provided by call centres. Not only is this easier for organisations, it’s also faster and cheaper. Q-go’s headquarters are located in Amsterdam, and the company has four local offices in Barcelona, Madrid, Frankfurt and Zurich.

A post by Don Dodge (MSFT) reveals that the Powerset’s “semantic rules” can be applied to MSFT’s existing index of the web and therefore likely to be as much about word word co-occurence statistics + clusters as it is using linguistic logic to gain an understanding from the corpus. The example is also illustrative of how Powerset breaks down a natural language query, and the post goes onto explain how Powerset may also be useful in vertical search applications…

Powerset is using linguistics and (NLP) to better understand the meaning and context of search queries. But the real power of Powerset is applied to the search index, not the query. The index of billions of web pages is indexed in the traditional way. The big difference is in the post processing of the index. They analyze the indexed pages for “semantics”, context, meaning, similar words, and categories. They add all of this contextual meta data to the search index so that search queries can find better results.

Who is the best ballplayer of all time? Powerset breaks this query down very carefully using linguistic ontologies and all sorts of proprietary rules. For example, they know that “ballplayer” can mean Sports. Sports can be separated into categories that involve a “Ball”. Things like baseball, basketball, soccer, and football. Note that soccer does not include the word ball, yet Powerset knows this is a sport that includes a ball.

Powerset knows that “ballplayer” can mean an individual player of a sport that includes a ball. They know that “best of all time” means history, not time in the clock sense.

Finally, an alternate theory that says Powerset was acquired for contextual ad placement, not search.

For the foreseeable future, phrases and statistical approaches will probably continue to deliver the greatest signal-to-noise-ratio for machine indexing of web content in the absence of a breakthrough in artificial intelligence. The evidence is not 100% conclusive, but during the last several weeks I’ve accumulated research papers to support this hypothesis…

• Part-of-speech (POS) tagging for noun phrases is the standard way to find collocations, an approach discussed here based on statistics may work better: this is based on late-1980s work by Church and Hanks “Word Association Norms, Mutual Information, and Lexicography” and “Using statistics in lexical analysis.”
• For a textbook introduction, see Chapter 5, “Collocations” and Chapter 7, “Word Sense Disambiguation” in Foundations of Statistical Natural Language Processing by Christopher D. Manning, Hinrich Schuetze
  
• By constructing a similarity measure between similar words in n-grams (n<=5) using the Wikipedia corpus, the authors of “Finding Similar RSS News Articles Using Correlation Based Phrase Matching” found that they could discover related articles and the highest accuracy was achieved for bigrams (91%) and trigrams (90%) versus unigrams (86%), 4-grams (70%) and 5-grams (60%).
• The paper “More Effective Web Search Using Bigrams and Trigrams” found that phrases (bigrams and trigrams) found by analysis of the top 100 initial search results in typical2-3 word search queries (1) improved the relevance of subsequent searches that included the bigrams/trigrams (2) did even better after verifying the bigrams/trigrams with the user (using relevance feedback) and (3) improved readability of topic areas beyond the first ten results.
• “A Language Model Approach to Keyphrase Extraction” by Tomokiyo compares metrics to rank phrases in terms of their informativeness, and finds derivatives of a KL-metic win over likelihood ratios – this kind of phrase-finding technology was used at Intelliseek in “BlogPulse: Automated Trend Discovery for Weblogs” and “Deriving marketing intelligence from online discussion”.
• The authors of “The Use of Bigrams to Enhance Text Categorization” found that bigrams improved the quality of feature sets used in Naïve Bayes classification of documents into categories.
• For a given word, assignment of the single sense per phrase containing that word that appears most commonly in a document was shown to be highly effective (90-97% accuracy) for word-sense disambiguation in “Unsupervised Word Sense Disambiguation Rivaling Supervised Methods” by David Yarowsky – also see pages 250-253 on “Foundations of Statistical Natural Language Processing” by Manning/Schuetze.
  
• In “Diverse Topic Phrase Extraction through Latent Semantic Analysis,” … “We propose a novel algorithm for extracting diverse topic phrases in order to provide summary for large corpora. Previous works often ignore the importance of diversity and thus extract phrases crowded on some hot topics while failing to cover other less obvious but important topics. We solve this problem through document re-weighting and phrase diversification by using latent semantic analysis (LSA).”
• New word senses can be discovered based on representing co-occurrence and using SVD in “Automatic Word Sense Discrimination” by Hinrich Schutze.
• Phrase-clustering of clickstream content to detect user interests (Slovakia, 2002). This book chapter (Intelligent Support for Information Retrieval in the WWW Environment, 2002) uses clustering to automatically infer/detect a user’s interests from inside the content of web-pages that the user visited – personalization derived from the clickstream content. Reports an 80% success rate in classification using unigrams, bigrams, and trigrams.

Also some papers on techniques/tools for extracting keyphrases and performance evaluations that I came across…

• Google made n-gram data available for all 5-grams (or smaller) appearing more than 40 times on the web, “All Our N-gram are Belong to You”
• Open-Calais by Reuters can analyze for named entities, facts, and events and their API
• The chapter “Natural Language Tools” in pages 149-171 in “Advanced Perl Programming” by Simon Cozens (O’Reilly) — you get a very “quick & dirty” introduction to a number of natural language processing concepts and ways to implement and play around with them. Although Perl has many natural language processing tools, the Cozens book cuts to the chase, explains which are the easiest tools to use, and shows you how to use them.
• “Coherent Keyphrase Extraction via Web Mining” discusses four variants of the KEA key-phrase extraction algorithms – two baseline and two enhanced – and evaluates their performance on a set of computer science and physics research papers. The enhanced versions use statistical association based on web term/phrase frequencies to “weed out” key-phrases that are unlikely to be associated with the text, and they improve on the baseline by generating a higher percentage of key-phrases that match author-generated key-phrases.
• “Narrative Text Classification for Automatic Key Phrase Extraction in Web Document Corpora” compares three key-phrase extraction algorithms (TF.IDF, KEA, and Keyterm) for summarizing web-pages on websites. They find that Keyterm performs best in tests with users. KEA is a keyphrase extraction package, and another one is Lingpipe.
• Paper from MSFT Research for clustering algorithms to find user-verifable “related” phrases within search results — could in theory be applied to any list of documents, “Learning to Cluster Web Search Results”
  
• Autonomy’s Clustering for Enterprise.

Millennium Technology Prize

Millennium Technology Prize Awarded to Professor Robert Langer for Intelligent Drug Delivery

The youtube video on this site interviewing Dr. Langer is cool… he shows how drug delivery via polymers is now leading to precise targeting of drugs down to the unicellular level and enabling release of drugs controlled by human-embedded microprocessors.

In choosing his career in 1974, he blew off the oil companies and talks about how his first boss liked to hire unusual people. He invented 200 ways that it didn’t work for every 1-2 successful ways that did.

Fermi’s nobel lecture (1938)

The simplicity of Fermi’s nobel lecture (1938) is stunning — the implications of this work changed history forever. Other nobel lectures i’ve read go on and on — this lecture is only 8 pages. Fermi also cites and gives credit to a dozens of other researchers upon whose work his discoveries are based. He explains the discovery of radioactivity caused by neutron bombardment and study of interactions of “thermal” neutrons with all the elements, including uranium and thorium.

p. 415,

The small dimensions, the perfect steadiness and the utmost simplicity are, however, sometimes very useful features of the radon + beryllium sources.

His experiments involve neutron sources, paraffin wax, and spinning wheels, not complicated particle accelerators or machinery. Anyone with a freshman-level chemistry/physics knowledge should be able to understand the lecture, but even that is not absolutely needed.

Seth Klarman on the credit crisis and value investing

This guy has delivered 20% net compound returns for 25 years.

“The pendulum may be starting to turn-as recent developments in the mortgage and hedge fund markets suggest. Because the scale of today’s leverage so greatly exceeds historical levels, it seems possible that we are only in the early stages of a credit contraction.”

His speech at MIT in January below covers a lot of ground on value investing versus leverage, efficient markets, credit crunch, VC/PE.

The Benefits of a Long Childhood

In his essay “The Benefits of a Long Childhood,” Ethan Remmel reviews “Why Youth Is Not Wasted on the Young: Immaturity in Human Development” by David F. Bjorklund.

Children’s thinking differs from that of adults, and people tend to view those differences as deficits that need to be overcome-the sooner the better. Bjorklund argues, though, that some aspects of children’s immature cognition are actually adaptive, both in preparing them for adulthood and in allowing them to flourish in childhood. He gives several examples: Children typically overestimate their own abilities, which may maintain their motivation in the face of failure and lead to eventual success. Their limited information-processing capacity may help them learn language, because it forces them to focus on constituent components and build upward from there, whereas adult language learners skip straight to semantics, often failing to master underlying grammatical structures. And play in childhood may promote later social competence, as neuroscientist Sergio Pellis has demonstrated in rats.

Bjorklund finds implications that will interest parents and educators. Parents often want their children to be the first among their peers to reach every developmental milestone, but Bjorklund points out that earlier is not always better and may sometimes be worse.

For example, abnormally early visual experience in birds disrupts development of the auditory system. Also, in a classic study published in American Scientist in 1959, “The Development of Learning in the Rhesus Monkey,” psychologist Harry Harlow found that the ability of rhesus monkeys to discriminate objects on various dimensions such as shape was impaired by starting the training too early in life—the monkeys who started training at older ages reached higher peak levels of performance. In a 1977 study by developmental psychologist Hanus Papousek, human infants who started learning to turn their heads to specific sounds at 31 days of age mastered the task, on average, at 71 days of age, whereas infants who started learning to do so at birth did not master the task, on average, until the age of 128 days.

Bjorklund’s message is that human development takes as long as it does for good reasons and that experiences should be introduced only when children are cognitively ready for them. Early education should foster a love of learning, which will pay dividends in the long run, rather than a fear of falling behind, which increases stress and decreases motivation. He acknowledges that schooling is necessary for success in the modern world and that direct instruction is sometimes useful. But as much as possible, he believes, we should let children enjoy childhood. We should even seek to maintain some “immature” qualities, such as curiosity and playfulness, into adulthood. As Aldous Huxley observed, “The secret of genius is to carry the spirit of the child into old age, which means never losing your enthusiasm.”

This may be a contributing factor to why Finnish kids by age of 15 — who start school 2-3 years later at age 7 than students in other countries — outperform their peers on international standardized tests conducted by the OECD.

Book Review of “The White Man’s Burden”

On a long plane plane ride to India, I read the “The White Man’s Burden: Why the West’s Efforts to Aid the Rest Have Done So Much Ill and So Little Good ” by William Easterly – the guy that Bill Gates had a strong reaction to recently on a panel discussion in Davos earlier this year. Thanks to Mora for suggesting and lending the book to me.

From “Bill Gates Issues Call For Kinder Capitalism,” January 24, 2008, Wall Street Journal…

“If we can spend the early decades of the 21st century finding approaches that meet the needs of the poor in ways that generate profits for business, we will have found a sustainable way to reduce poverty in the world,” Mr. Gates plans to say….

To a degree, Mr. Gates’s speech is an answer to critics of rich-country efforts to help the poor. One perennial critic is Mr. Easterly, the New York University professor, whose 2006 book, “The White Man’s Burden,” found little evidence of benefit from the $2.3 trillion given in foreign aid over the past five decades.

Mr. Gates said he hated the book. His feelings surfaced in January 2007 during a Davos panel discussion with Mr. Easterly, Liberian President Ellen Johnson Sirleaf and then-World Bank chief Paul Wolfowitz. To a packed room of Davos attendees, Mr. Easterly noted that all the aid given to Africa over the years has failed to stimulate economic growth on the continent. Mr. Gates, his voice rising, snapped back that there are measures of success other than economic growth — such as rising literacy rates or lives saved through smallpox vaccines. “I don’t promise that when a kid lives it will cause a GNP increase,” he quipped. “I think life has value.”

Brushing off Mr. Gates’s comments, Mr. Easterly responds, “The vested interests in aid are so powerful they resist change and they ignore criticism. It is so good to try to help the poor but there is this feeling that [philanthropists] should be immune from criticism.”

also see Easterly’s rebuttal.

Easterly is former research economist at the World Bank now at NYU. in his book he looks at the successes/ failures of international aid interventions (financial + military) by “The West” and makes the case that they have done more harm than good during the past 50+ years.

Most of Easterly’s book makes sense to me and I agree with Easterly that philanthropic/aid agencies are not “above” criticism – their hyped up expectations do not necessarily make things better and sometimes they make things worse by standing in the way of more realistic, lasting solutions… but,

1. I agree with Gates on one thing, that you can get into trouble measuring national economic development using aggregate GDP (growth) instead of measuring the purchasing power of the bottom pyramid (half or quarter) of earners in the economy. Easterly cites India as a success story of development showing a chart of exponential GDP growth over 20-40 years using the Indian IT industry as an example. Despite this progress, the failure is that 50 years after Indian independence close to half of all Indians (400-500 million people) still live on less than $1 of $2 per day.
2. Easterly says new (niche) market creation is limited by social and legal barriers to trust and property rights and therefore must take place indigenously – there is not much we can do about it living in “The West.” I think we have not yet explored the potential of the Internet to overcome these constraints and help diversify agricultural economies (long tail). See my essay “Opening Niche Markets in Rural India using the Internet”
3. Style. Though his analysis is very compelling and data-driven with graphs, stories of people, case studies of developing nations, and world history, to me the title seems a bit polarizing or stuck in the past and the tone of the writing is funny bit also feels a bit sarcastic. Perhaps it is discouraging to visionaries and optimists who want to break from the past. In his book he takes aim at Bono and Jeffery Sachs’ “The End of Poverty.”

I tried to capture the main ideas of the book… sure I missed something but I think it’s mostly here.

— Top-down “planners” at large institutions like the World Bank will mobilize resources on the basis of utopian agendas and large-scale “big pushes” that attract donor governments and private institutions (in US, UK, and The West). These visions are never achieved because they lack feedback from the people (Africa, Asia, and “The Rest”) whom they are intended to benefit, i.e. the poor. On the other hand, he notes that the World Bank produces very high quality economic research.

— Unlike market-driven firms or (legitimately) elected officials the planners are accountable to donors, not the poor. planners’ jobs are not dependent on serving the poor but rather to indulge donors’ unrealistic expectations which may never materialize. The “planners” efforts do more harm than good (large part of what the book is about). The failures of these big pushes become self-fulfilling as donors redouble their efforts, bureaucracy becomes bloated, and they begin to measure progress based on volume of aid disbursed not impact on the poor. incentives of planners and poor people are not sufficiently aligned – this is called the principal-agent problem

— Bottom-up “searchers” (NGOs, entrepreneurs, profit-seeking companies) who are on “the ground” in developing countries can get direct feedback from the poor people they serve and make real impact on the their lives. they set realistic, achievable goals unlike the planners. Too little money is going to support the searchers. several case studies.

— On microfinance and microcredit,

Microcredit is not a panacea for poverty reduction that some have made it out to be after Yunis’ discovery. Some disillusionment with microcredit has already come in response to these blown-up expectations. Microcredit didn’t solve everything; it just solved one particular problem under on particular set of circumstances-the poor’s lack of access to credit except at usurious rates from moneylenders.

— Markets are a spontaneous outgrowth of social trust (for transactions) and property rights (for investment), and can’t be planned by aid agencies, foreign governments, or “out of the blue” after an invasion or removal of a dictator.

— Foreign aid has been most effective and made a large-scale impact in people’s lives for things like vaccination, health care delivery, and programs to keep kids/girls in school when compared to other areas in which results can’t be directly measured. dollar for dollar, the recent momentum to offer AIDS treatment ($1000 per person) like Bush’s $15 billion commitment of US taxpayer funds for Africa (30 million infected) is many times less cost-effective compared to preventing the spread of AIDS through condoms (600 million not infected) or even prevention of other life-threatening diseases like malaria, diarrhea, and infant mortality. the spread of AIDS could have been avoided had prevention been a bigger priority since experts have been predicting this epidemic for over a few decades. In AIDS, saving a life gets more “emotional” attention from the public than prevention of AIDS transmission which could save many more lives.

— There have been some success stories, but economic growth in developing countries has not been correlated to aid/intervention by the West. Colonialism and imperialism has resulted in long-term economic stagnation, which he offers as a case study to consider other neo-imperialistic plans to take over weak-states. His claim is that countries develop much faster and better when they are left to their own.

— National financial health has less direct impact on earnings of the poorest people, except indirectly via inflation and government subsidies to the poor.

The IMF’s approach is simple. A poor country runs out of money when its central bank runs out of dollars. The central bank needs an adequate supply of dollars for two reasons. First, so that residents of the poor country who want to buy foreign goods can change their domestic money (let’s call it pesos) into dollars. Second, so those poor-country residents, firms, or governments who owe money to foreigners can change their pesos into dollars with which to make debt repayments to their foreign creditors. What makes the central bank run out of dollars? The central bank not only holds the nation’s official supply of dollars (foreign exchange reserves), it also makes loans to the government [aside from foreign borrowing with bonds] and supplies the domestic currency for the nation’s economy. The government spends the currency [it borrows], and the pesos pass into the hands of people throughout the economy. But are people willing to hold the currency? The printing of more currency [excessive government borrowing from the Central Bank] drives down the value of currency if people spend it on the existing amount of goods – too much currency chasing too few goods… so they take the pesos back and exchange them for dollars. The effect of printing more currency that people don’t want is to run down the central bank’s dollar holdings. Too few dollars for the outstanding stock of pesos is kind of like the Titanic with too few lifeboats. The country then calls on the IMF. So the standard IMF prescription is to force contraction of central bank credit the government, which requires a reduction in the government’s budget deficit [government spending]… forces the government to do unpopular things [like cut subsidies] – disturbance of domestic politics.

— Bad governments (corruption and violent dictators) have been responsible for much of the slow growth in these countries, which are in turn caused by either a colonial past or by historical poverty itself. Foreign aid tends to prop these governments up, and in some cases private organizations working around these governments can lead to much better results.

— Loans are not necessary to balance a national budget, and the IMF’s prescriptions for foreign exchange lending to developing countries and reducing government spending can be way off. This is due to severe accounting irregularities in the books of these countries, uncertainty of how or when markets react to falling currency prices, and how they react to information in the economy (people’s behavior). Lending based on shaky foundations can lead to the self-reinforcing “debt trap” through repeated refinancing of poor countries and propping up of bad governments. the IMF does better in emerging markets, but he says it may be better off to leave the poorest countries alone.

Kiva: The Ebay for Microfinance

Yesterday I went to the talk by Premal Shah about www.kiva.org which he calls the “Ebay for microfinance.” It was at Zerox PARC. Thanks to Chari for giving me a heads up about this. They get lots of press and also see a recent NY Times article about them.

 

Kiva is a nonprofit website and clearing house that enables Internet users (lenders) to give 0% (interest-free) loans to specific individual or small-group “entrepreneurs” (borrowers) in developing countries in Africa, South America, Eastern Europe, and Asia. They work with 85 field partners from 40 countries, called microfinance institutions (MFI), who post/vet entrepreneur profiles which are in turn selected by Internet users for personal loans. One of the biggest advantages of Kiva is end-to-end transparency — each lender can “see” who which borrower they are lending to, track their progress through journal updates, and see when the loan is being repayed. See a blog post by Guy Kawasaki that explains their fee per transaction business model — $2.50 voluntary fee that lenders pay when checking out their “shopping cart.”  

Yunis was first 30 years ago, and today there are 10,000 MFIs worldwide. He estimates there are 500M people needing loans like this, and only 100M have been reached through traditional microfinance to-date. Access to capital is still a bottleneck he says. Note: Kiva is prevented from operating in India due to their bank regulations.

Kiva Statistics. Kiva is 3 years old, so far $20M loaned by Q1 08 in 3 years since inception,. Observing parabolic quarter on quarter growth in loans and expect to have loaned $250M to $1B in five years surpassing efforts of microfinance initiatives major banks like Citibank ($100M).  

The Kiva model:  

Internet user (a lender, social investor) 

–> Kiva (online marketplace)  

–> Local Field Partner (Microfinance Institution)  

–> Developing World Entrepreneur (the borrower) 

Each Kiva lender has given on average 2.2 loans — $25 max limited by Kiva. Over the 20-30,000 people visit the site daily, and 3% end up giving loans. Each borrower is usually funded by 15-20 lenders on average — typically in the range of $1000 total loan size. The lenders usually sign up within a day of a loan request being posted for an entrepreneur. In some war torn or crisis areas like Iraq of Afghanistan, the lenders sign up in just a few hours. 

The lenders (social investor) rationale is that Kiva is transparent (know where it goes), sustainable (if repaid, money can be lent to someone else), affordable ($25 to change someone’s life), and unique (“I love microfinance, I want to participate”). For the microfinance institution, Kiva offers a low interest US dollar capital, no liability, flexible repayment terms, and financial assistance + incentives for transparency. 100% of loan funds go directly to the borrowers.

Kiva checks out (verify) the MFIs and the MFIs in turn check out the borrowers. The MFI charges 20% interest rate to cover distribution costs, and they bear currency ris as well when converting from US dollars to local currency and back during repayment period. In the future Premal hopes they can establish credit histories for individuals and bypass this intermediate layer completely. Kiva uses random sampling to audit and check on the MFIs for fraud.  

Unlike traditional microfinance where borrowers are organized into groups who are accountable to each other, social accountability is created via the Internet. Borrowers’ profiles are made visible on the Internet, and is therefore visible online to locals who watch each other at Internet kiosks/cafes. 

Capital. Internet users are willing to bear greater risk than banks probably due to the “personal connection.” People don’t want (need) to lend with interest, whereas banks have to when they are working with MFIs.  

Operating Principles. “Unreliable credit is OK, but unreliable data is not OK.” Each lender gets a “portfolio” of people who they lend to, much like a stock portfolio on Etrade. Loans in this model create a “persistent tie” between the people across the world with journal updates. The accountability is simple. If you are getting repaid, something is working. If you don’t get repayed, something is not working. MFIs report repayment rates of 99.7% but Premal believes that’s skewed to report better than actual results because the MFIs don’t want to discourage Kiva. He believes the actual repayment rate is over 90%. There is a “Risk & Due Diligence Center” on the website. 

Diversification. I asked what types of different activities and market opportunities are being funded, epecially outside agriculture. He said he doesn’t know for sure, but says that agricultural opportunities are the dominant activity being funded. In addition, there appears to be a slight lender bias. For example, the most popular kind of loan that gets funded by lenders is African female farmer, and the least popular is an Eastern European male taxi driver. He questions whether or not that is economically rational, but says that’s what it is right now. In 3Q 2008 they plan to open up APIs so researchers can download and analyze the loan data from their website to gain further insights. 

Organization. Kiva.org is 25 people and operates on very low overhead thanks to cooperation and donations from several silicon valley companies. For example PayPal gives free payment processing. Google offers free AdWords traffic.  

Spaceship One, Spaceship Two

According to news this week, for $200K you will be able to fly up into space in a few years, “Entrepreneur Unveils New Tourist Spacecraft”

On October 8, 2006 in San Carlos, CA at the Hillier Aviation Museum, I had the good fortune of listening to Burt Rutan speak about breakthrough innovation, aviation, spaceflight, and aviation safety—totally inspiring. Some of it is very big picture, but here are a few of the highlights,

• Technical progress and the ability to take big risks has been what sets humans apart from animals
• Children make the decision to be innovators during the age of 3-14, usually due to some events that occur during that period.
  • Most of the aviation pioneers that people recall (Von Braun, etc.) were growing up during the time when the airplane was invented early 1900. Most aviation since then has used the same basic principles they discovered
• Next major wave of innovation occurred WWII and after when he was growing up.
• Third wave was when Sputnik made Americans feel they lost to the Russians, which kick started the space race of 1960s.
• Since then, we have been using essentially the same space technology for last 30 years.
• His SpaceShip One is now housed in the Smithsonian right next to the other major plans of the last century
• Commercial & military airplanes have stagnated in their altitude and speed because the technologies have not been pushed by the organizations that develop them. Space Ship One pushes the envelope by orders of magnitude, representing the next wave of innovation in aviation and space travel. Space Ship Two will be for commercial travel—Virgin airlines may be accepting orders for the first flights.
• Safety and stability have been the barriers to entry in commercial space flight—that’s what he’s out to change. One of his first jobs out of college was to understand why the F4 had so many failed flights and engineer a stability control system.

Below is a photo of him telling me what he thinks about the Columbia accident report: “if you read it carefully, what they are saying is not to take risks. NASA as an organization will never take risks.” Also asked him what he thinks is the difference between his small 130 person company and NASA is. He replied that he never puts his engineers and factory personnel in the position of defending safety, i.e. never to be in a defensive position, or allow an aviation regulator do that to them.

• I read this to mean this places full responsibility in the people doing the work to ensure safety.
• Safety has to be so obvious to the people doing the work that there is never a need to be defensive—they understand exactly why their aircraft is safe.
• My interpretation is that this nurtures a culture which outperforms regulated safety—he claims he has built some 40 research aircraft with an excellent safety record he claims

Also see, “Larry Page on how to change the world: Breakthrough ideas are around the corner, says the Google co-founder.  But most of us are failing to take a chance on them.”

Trust-based lending by the inventor of microcredit

This week we got to see the founder of Grameen Bank, Mohammed Yunnis speak in person at the Fairmont Hotel in downtown SF (Commonwealth Club). He pioneered a new form of banking that can be called “trust-based” or “community-based” lending, which is contrast to the worldwide banking system based on “collateral” and “credit-histories.” For this he won the Nobel Peace Prize in 2006. See this FAQ http://www.grameen-info.org/bank/GBGlance.htm

In 1976, he was a economics professor an Dhaka university where he found the theories he was teaching had little practical applicability to the poverty he was seeing outside the university. So he decided to do what he could. He lent small amounts of money (~$500) to poor people and they became very happy. Then he asked, why doesn’t the bank do this?

For months he struggled to get banks to lend to the poor, but found them to be uncooperative. Since poor people didn’t have credit histories + collateral the banks were unwilling to lend money. His insight was to turn those assumptions on their head by lending to people with no money and history of taking loans. He sought out poor women who were afraid of taking money, and tells his employees that those are the people they should loan to.

No lawyers. By challenging all the “assumptions” he came up with something completely new. He said that the current banking system has lots of money and is setup to loan large amounts of money to people who already have money. That architecture doesn’t scale down to vast majority of people who need it. Using the analogy of ships, he said the modern banking system is like a large supertanker. For poor people you need a system more like a dinghy boat, and if you scale down a supertanker architecture to that size it would sink.

Grameen Bank does not require any collateral against its micro-loans. Since the bank does not wish to take any borrower to the court of law in case of non-repayment, it does not require the borrowers to sign any legal instrument.

Although each borrower must belong to a five-member group, the group is not required to give any guarantee for a loan to its member. Repayment responsibility solely rests on the individual borrower, while the group and the centre oversee that everyone behaves in a responsible way and none gets into repayment problem. There is no form of joint liability, i.e. group members are not responsible to pay on behalf of a defaulting member.

Social business. Yunis described the concept of “social business” intended simply to help people, which can exist alongside profit-maximizing businesses and often work in synergy.  Related, see January 24 front page article on Gates speech at Davos, “Bill Gates Issues Call For Kinder Capitalism”

“If we can spend the early decades of the 21st century finding approaches that meet the needs of the poor in ways that generate profits for business, we will have found a sustainable way to reduce poverty in the world,” Mr. Gates plans to say.

Everyone is an entrepreneur. Yunis explained that all people are entrepreneurial by nature, and it is the system that either brings it out in them or not. Grameen Bank has been able to convert tens of thousands of beggars into door-to-door salespeople who earn a living – since they visit those house anyways, why don’t they take along something to sell and make money? It turns out beggars have unique knowledge of when particular people are home, and and household demographics that is useful in selling.

Begging is the last resort for survival for a poor person, unless he/she turns into crime or other forms of illegal activities. Among the beggars there are disabled, blind, and retarded people, as well as old people with ill health. Grameen Bank has taken up a special programme, called Struggling Members Programme, to reach out to the beggars. About 98,500 beggars have already joined the programme. Total amount disbursed stands at Tk. 102.27 million. Of that amount of Tk. 69.74 million has already been paid off

There are many other programs (scholarships, cell phones, loan insurance, etc) described on the website. Some facts,

• Total amount of loan disbursed by Grameen Bank is US $ 6.55 billion
• Loan recovery rate is 98.35 per cent
• Total number of borrowers is 7.34 million, 97 per cent of them are women
• Grameen Bank finances 100 per cent of its outstanding loan from its deposits. Over 58 per cent of its deposits come from bank’s own borrowers. Deposits amount to 139 per cent of the outstanding loans
• Ever since Grameen Bank came into being, it has made profit every year except in 1983, 1991, and 1992.
• Grameen Bank has 2,468 branches. It works in 80,257 villages. Total staff is 24,703.

I expect the Commonwealth Club audio transcript should be available in a few weeks. Here are some funny short clips from where is interviewed on the Daily Show (2006) right after recieving his nobel prize where he uses the phrase “trust-based lending.” He was also interviewed a few weeks ago by Steve Colbert (2008). Speaking to Colbert, Yunis remarks that the current home loan default crisis in the US was caused not by the people who took the money but rather by the banks who didn’t understand how to lend money.