Carson Block Interview (4/15)

Tencent recently published an interview conducted with famed US short seller Carson Block of Muddy Waters Research on April 15th, 2020. The interview covers a lot of ground, including several of Block’s current and former positions (Luckin, IQiyi, Anta, TAL), short selling strategies, China capital flows, and teambuilding.

*While the article and transcription were published in Chinese, the interview itself was done in English. This is my transcription (lightly edited for grammar and fillers).

For more fun stuff, you can follow me on twitter @kevg1412. And definitely check out my compilations page.

Some people say that Snow Lake Capital is your partner is China. Is that an accurate description?

We don’t have partners in China, but I certainly know who Snow Lake is. I’ve talked to the principal a few times over the years, but we don’t have a partnership by any means. We don’t have a partnership with anybody in China.

Before reaching short sellers, the report was sent to shareholders on the secondary market. But that didn’t move the stock price. Did that particular make Luckin a very difficult short target for you?

So I think there’s a hidden question here, right? You’re asking me who’s behind the Luckin report? That’s really your question. I’m not going to comment on who’s really behind the Luckin report. But I will tell you that I knew who the author was, so in my tweet that I sent out I referred to the report as unattributed as opposed to anonymous. So I knew who the author was. It’s somebody whom I’ve known for some years, and I felt is very credible in the space. It’s somebody who I found was very measured in what he said.

We got access to a data room behind this report, so we were able to do a sampling in order to validate the work. And then there were aspects of this report, I mean, on the cover page it talked about how the author stated that he suspected that one of the ways they round tripped money was by working with Focus Media. So look, anytime you want to talk corruption, you have me at Focus Media. We exposed them in 2011. They were able to go private. Carlyle and another PE firm, Fountainvest, drove the getaway car. But they still after going private, settled with the SEC for about $56MM. So, I think that’s an admission of something. But in any event, you have me at Focus Media.

Yes, but my question is that other shareholders also got the report even before you. And actually, after your tweet on January 31st, their stock price still went up. So what did you do then?

Okay, so, a couple of things. I’d never tweeted out somebody else’s report and saying we’re short. But we knew that a lot of other people had this report. We figured it’s only a matter of time before it gets out, and we also thought we were up against the clock, so to speak, because we thought it would get out anyway. So we made the decision to put it out ourselves.

Now, why didn’t the stock continue to go down? If you look at most of the page one holders, and this goes back to what I was saying earlier. I think they felt that the report was correct. They just felt it didn’t matter. You’ll have these large, sophisticated hedge funds say, “Yeah. It’s a fraud. It’s a big fraud. But whatever. We should still own it.”

After January 31st, the stock price remained stable. What did you do back then? Did you keep building or  selling your position?

No. Most of our positions we now try to beta hedge. We had a very manageable position, and wanted to see what would happen. The way that short activism works when it’s successful, the analogy I’ve often used is, you’re knocking over a bunch of dominoes when you begin your campaign. How many dominoes will continue to fall, which dominoes will continue to fall, is very difficult to predict. But a lot of times, the vast majority of times, when we go activist on something, a lot of dominoes do fall. And things inside the company start to break. And those things that break can be, maybe the auditor decides to do its job, which in this case, is what I think happened. Maybe board members or managers get uncomfortable and they won’t continue to commit fraud or do things that are very aggressive.

So, things broke. But we don’t, especially when it’s not our report. The only thing that was really, that factored into how we ran the position was, “It’s not our campaign.” It wasn’t clear to me that the author of the Luckin report would do anything to follow up. So that was one risk factor that factored into how we ran the position.

Did you expect the size of the fraud to be so big? $300MM?

Oh. [chuckles]. Yeah. [chuckles]. And I think there’s more to what actually happened than what’s been released. I don’t believe that it’s one guy. That doesn’t hold water. And I suspect it’s larger than that in terms of the revenue.

What is a derivative of a company that’s committing fraud at the most senior levels of management worth? I think zero. But like I said, other people might have different views there.

Do you think David Li and Er-Hai Liu were both well aware of the big fraud of the company?

They probably are the big problem. [chuckles]. That’s my view.

20 years ago, with the Enron scandal, one of the big five auditing firms disappeared. Do you think history repeats again?

No. So first of all, two things changed since then.

Number one: It seems like immediately after that happened, the US government and policy makers came to feel that it was a bad idea to put an entire auditor out of business, or an entire firm out of business. And that’s one of the reasons why enforcement since the Enron days has been pretty loose. That’s number one.

Number two. What the auditors did was, they set up what I really think are legal fictions, but they’re now structured as these networks of independent entities. And look, I hate the auditing profession. I think it has absolutely bamboozled investors and continues to do so. Because up until the moment something goes wrong, Partners move, people move freely from one entity to another. They’re seconded. And money moves freely. And the global partners will make money off of these local affiliates. The moment something goes wrong though, the story totally changes. “Oh, no. That was our local affiliate. You can’t hold us in the US or in the UK, the global groups, you can’t hold us responsible.”

At the end of the day, you have to blame investors, right? Every year, every one of the big four will have a major blowup. And yet everybody thinks the big four brand means something. Every time we short a company, especially if we say it’s a fraud, I just see all of the responses on Twitter. And it’s all “Oh, but it’s audited by so and so.” Like, come on man. So and so last year had three major audit failures.

So anyway, the audit profession, it will continue, it will be business as usual. I think nothing will change in that respect.

You just mentioned the auditor of Luckin. There were also other Tier 1 agencies on the deal, like Credit Suisse, Morgan Stanley, CICC, Haitong. Are they also to be blamed? What’s their role in the whole thing? What will happen to them after this?

They’ll settle for some money. It’s a cost of doing business.

What’s their role? Years ago, right after we published Sino-Forest, I spoke with the credit analyst at one of Sino-Forest’s investment bankers. This analyst told me, “I figured out that Sino-Forest was a fraud over a year ago. I went to my boss, and I said that.” And my boss said, “Listen, you just be quiet. You don’t have to cover the company.” But, that’s it.

So on the equity side of the house, that firm, which is one of the major firms in the world, maintained a strong buy on the equity. That firm continued to do underwriting business for Sino-Forest. They don’t care. Caring is bad for business. They’ll pay settlements every now and then.

It’s kind of funny. I had a discussion several years ago with somebody at a major investment bank. We’d been talking to this investment bank about a prime brokerage relationship. They said, “I’m getting some pushback here because what you guys do, it’s sharp-elbowed, it’s kind of edgy.” I said, “You’re sitting in the headquarters in New York right now. You know at this minute, this very moment, somebody in that building is doing something that is going to lead to at least a $50MM penalty for the bank somewhere down the line.” He went, “Yeah. Yeah. It’s true.” But that’s the thing. Banks look at it as a cost of doing business when they occasionally have to settle for something that they do, like, in underwriting.

I don’t want to ask this question myself, but as you know, a lot of my friends always turn to me for this. Do you think Starbucks is behind the short selling of Luckin?

[Laughs out loud]. You know, every single time I shorted a Chinese company, that’s the defense. “Oh, it was the competitor.” No. I don’t think so. I don’t think Starbucks cared. [chuckles]. I don’t think so.

To what degree did you participate in Wolfpack’s investigation into IQiyi?

We did a decent portion of the research, and Wolfpack did some of the research. Wolfpack ended up directing a good portion of the research. We supported what they wanted us to do in certain areas. We were working on that one for literally a little more than a year.

The report of IQiyi only samples only 613 persons in Tier 1 cities, far from the 107MM members they published. Do you think it’s an appropriate sample size?

Actually, no. There were over 1500, approximately 1600 people sampled in Tier 1 cities. That might be referring to the percentage who responded that they’re VIP members.

So we ended up bringing in a statistician to help us evaluate whether the results were statistically significant. That’s something to which we were very sensitive. So there were certain data points where the sample size wasn’t big enough to draw a conclusion. There were certain questions where the sample size wasn’t big enough. But what we based those conclusions in the Wolfpack report on, were confirmed to be statistically significant because we were getting very tight ranges regardless of the day, regardless of the city. There were certain questions to which the answers were very consistent.

And also the report quoted only former employees for illustration. Do you think this is solid evidence?

So, if you were to go public and say, “I spoke to a former employee of IQiyi or XYZ Company, and that’s why I think it’s a fraud,” then that would be very, very poorly sourced and supported. And it would be highly problematic.

But when we looked at IQiyi, there are so many data points, and so many arguments, that support the conclusion that it’s a fraud. So citing the conversation with a former employee, was really just to try to further illustrate this and how this works as opposed to saying, “Aha! Here’s the proof! This former employee said XYZ.”

IQiyi doesn’t have, it’s not inventing 90% of its subscriber numbers. It’s not inventing 90% of its revenue. But it’s well north of that 30% bar that we want to clear.

So putting aside, that’s not inventing everything. The work, to me, the number of arguments, the amount of proof, that this company is inventing a substantial portion of its users and revenue is overwhelming.

So to single out just this one reference, I think really downplays the significance of everything else in the report.

What do you think about the fundamentals of Chinese stocks right now? They are quite different from Chinese companies that got listed 10 years ago, right?

I can’t really opine on fundamentals as opposed to, say, fraud risk. Do we think the revenue is substantially more than 30% fake? If so, can we prove it? What resources are needed to provide it? That’s really what the decision points are.

Now, when you look at TAL Education, which we wrote on two years ago, that’s not one where we attacked the revenue. We attacked the profit. We took a different angle there. Because we found periods in which the profit was substantially fake.

Sometimes we will break out of that framework. Because when we looked at TAL, we thought, “Ok, this is priced for perfection. This is a stock that everybody who’s invested in China loves. Yes, they really have learning centers. Yes, they really educate students. But, they’re lying.”

It turns out that that doesn’t really matter that much to investors in a bull market. It was worth a try. But anyways, here we are.

Speaking of TAL, you didn’t win in each and every case, like TAL and ANTA. They are still favorite kids of Wall Street. And also Pinduoduo, one of the most successful IPOs in recent years. What do you think of those so-called “failed cases.”

Ok, so we didn’t do Pinduoduo. That was not us. So yeah, Anta and TAL.

I wouldn’t call TAL totally unsuccessful. It wasn’t a win. It wasn’t a loss. TAL went down, and then it came back up with beta. So when people say, “Oh, but TAL’s back up,” I say, “Well yeah, but if you look at these PRC stocks since then, it pretty much came up in line, so [trails off]” Anyway.

Anta… Anta was a big loss. What can I say. It’s a fraud. I mean, it’s a real company. They’ve done impressive things. But the operating margins aren’t real.

But I think part of the problem for us when we look at Anta, and also when we look at TAL, is that you have so much money that is allocated to the China space. So these are decisions that are made by large allocators on a macro basis. “We’re going to be X percent US, Y percent Europe, Z percent China.” Ok, so now they’ve got Z amount of money going into China.  Well, “we need things that are large and liquid.” So that crosses off a lot of names. “We don’t want to be in state-owned enterprises.”  So that crosses off even more names. Ok, “What are we left with?” You’re left with a handful of companies that are just going to, no matter how problematic those companies are, I shouldn’t say no matter how problematic they are, but absent a showing of enormous problems, I guess, they’re going to continue to receive capital flows, at least in the environment that we had been in where central banks are just globally trying to force up asset prices as their economic playbook.

All these years of stimulating asset prices had really left investors, and not just investors, I think everybody in society in a way, but had really left investors anesthetized to risk. In other words, it’s like your brain has risk sensors, and for investors they’ve just been dulled, if not totally switched off by all these years of stimulus.

So in an environment in which people are reminded that there are risks, whether those are risks of things crashing, or financial assets crashing, or risks of global pandemics, or risks of natural disasters, or whatever, in that environment, I don’t know that this will continue to work. But obviously, the central banks, especially the Fed, are doing everything they can to push asset prices and make everybody forget about risk.

So we’ll see, next time around, whether these stocks, like a TAL, like an Anta, can continue to attract capital flows.

Given it seems that Chinese companies bear more risks, do you spend more time researching Chinese companies vs US counterparts?

The dynamics that I just laid out make it riskier for a short seller, because we don’t think there’s an issue that we proved, we don’t think there’s an issue that we proved that TAL has committed fraud. We don’t think there’s an issue with Anta having committed fraud. But, “do those things matter?” is the question.

That’s golden time for short sellers.

[chuckles]. It’s never a good time, right? Because everybody hates us. When the market’s going up, everybody’s trying to squeeze us, and everybody complains about us. And Elon Musk thinks that we’re the scourge of the planet. And then when stocks are going down, everybody hates us and blames us. So it’s never a good time to be a short seller.

You do it not because you say, “What’s the best way for me to make money in the stock market?”

You do it because you say, “I just feel like I know too much. I just don’t believe in so many of these stories that people chase in the equities market.” And that’s why you go into short selling.

After IQiyi and Luckin, do you think there will be another wave of attack on Chinese companies?

It’s tough to say. Effectively what I was saying earlier is, I don’t think that the incidence rate of fraud has really stopped. Now, the different between companies today and companies back in 2010 to 2012 is that those companies were 90%+ fraudulent top lines. We find a few of those now, we shorted China Internet Financial Services a few years ago, CIFS. That thing was as much as zero as NQ mobile or any of those other names.

But by and large, let’s say the universe of companies from China committing fraud. I think it’s still large. The amount of fraud is not, in terms of the percentage of revenues, is not what it was.

So the question really becomes, “Do investors think it matters?” That’s what we’re trying to figure out. Even outside of the world of China fraud, this is a question that we have to constantly answer. Because when we look at our bread and butter type of shorts, highly misleading accounting, etc. The bar has gotten higher each year because investors are just more and more unconcerned, just more and more oblivious, or deliberately oblivious to, risk.

So we’ve had to find things that are more shocking, or things that we think are more shocking or are just more problematic each year.

What’s the company size of Muddy Waters?

We’re now up at 7 core members of the team. But not everybody is working on China-related names.

What kind of roles do they play on your team?

We do have a former auditor.

We have somebody who has spent most of his career in manufacturing and sourcing. So he really helps us understand how people and material move through businesses.

We have somebody whose background was with a hedge fund in Hong Kong where he did a lot of behavioral type analysis. So he continues to do a lot of that behavioral  analysis for us.

We have a junior financial analyst who is pretty, he’s a very sharp guy, but maybe the most traditional finance pedigree.

We also have somebody whose background is in restructuring and complex capital structures.

And then myself. My background is first investing, then law, then real world entrepreneurship, and back to investing.

For more fun stuff, you can follow me on twitter @kevg1412. And definitely check out my compilations page.

Bret Victor: Visual Graphics (How Many Households)

I’m a big fan of Bret Victor, and have recently decided to read everything on his website again. Here are my notes to his article “How Many Households.” You can read the original here.

Key Takeaways

  • Show the data, compare the data, let the data guide reader interactions
  • The essence of data graphics is visual comparisons, even for static ones. Even if completely static, make sure the graphic informs. Add interactivity –judiciously, powerfully, consistently –to subset data space according to readers’ interests
  • Don’t make interactivity a barrier to information

Full Notes

  • The essence of data graphics is visual comparisons
  • Data can’t be compared when there are millions of data points, but only one is displayed at any given time
  • Many charts (like the one immediately above) require users to interact with a chart to switch between data points, often times through a convoluted navigation scheme with no clear information hierarchy, peek-a-boo menus that appear and disappear, and “actions at a distance” where clicking a menu causes a far-away image to change, and clicking the image causes the menu to change
    • This type of confused navigation scheme is not even the main problem!
    • The main problem is that any type of navigation is even needed. Chart readers have to continually interact with these charts, and have to beg for every scrap of info
  • The essence of data graphics is visual comparisons
    • An excellent graphic offers its info generously and eagerly, helping and encouraging its reader to answer wide-ranging comparative questions at a glance
    • An excellent interactive graphic offers many opportunities to explore the data space more precisely, to slice and filter the data, to see the data from different specific perspectives
    • Interactivity should NOT be a barrier to information
  • Keep information design and interactive design simple and consistent
    • Graphics should provide information right off the bat, and that information should inform the readers’ interactions
    • Every interactive element should be informational (ie bar graphs that display data AND provide a means to filter data; graphs that show distribution and slice the data in specific ways ie constrain a graphic to a specific variable)
    • There should be no UI buttons/control to clutter the graphic
    • Each feature should be offered simply by making the graphs consistent with the other graphs (ie click to filter)
    • Each click should potentially reveal a large amount of new data; this makes each interaction powerful, meaningful, and worthwhile. No gratuitous actions; every click rewarded
  • Some charts use unusual forms of multiple selection (ie when a category choice also needs a quantity specified, like the one immediately above)
    • Solution: being informational and consistent
    • For the second chart (“And who lives with them?”), say we want to see the data for “any children,” but also want to see the data for “1 child,” “2 children,” etc. Show all of the quantities as bar graphs, with the distribution of children expanding out when the “any children” bar is clicked
    • Interaction-wise, this graphic simply assumes the reader wants to specify one child, and immediately selects it, with the other selections just a click away
  • Keep designs stable. If something appears after a click, keep it there. Don’t move anything around, don’t make anything disappear.
    • This lets the reader take in and grasp the layout of the information space, and quickly explore new ideas without “navigation” getting in the way.
    • “Progressive disclosure,” where charts are introduced one by one, is to prevent the reader from being immediately overwhelmed by the graphic and skipping it (best used for general audiences, not necessary for sophisticated/motivated readers)
  • Another powerful technique is filtering on rollover instead of click
    • This allows reader to skim the mouse over a graph and immediately see the data sliced in every way –a huge information payoff for little interaction (can be awkward to integrate with charts like multiple-selection charts, but in general a solid technique)

2001 MIT Tech Review: 10 Emerging Technologies That Will Change the World

In 2001, the MIT Technology Review published a special edition of Technology Review: 10 Emerging Technologies That Will Change the World. You can read the original publication here. These are my notes. Some additional takeaways at the end.

Since 2001, MIT Technology Review has done yearly Top 10 Emerging Technologies issues. Notes for the following years will follow. I also plan on separate posts for additional thoughts, updates since issue publication (on both the highlighted technologies and the mentioned researchers), and related topics (ie new breakthroughs, interesting startups, related research) for each publication.

1. Brain-Machine Interfaces

  • Miguel Nicolelis is a leader in a competitive and highly significant field, in which there are only about half-a-dozen teams pursuing the same goals: gaining a better understanding of how the mind works and using that knowledge to build implant systems that would make brain control of computers and other machines possible
    • Nicolelis, working at MIT’s Laboratory for Human and Machine Haptics scored an important first on the HBMI (hybrid brain-machine interface, a term Nicolelis coined) front: sending signals from individual neurons in a monkey to a robot, which used the data to mimic the monkey’s arm movements in real time
      • Monkey has sockets installed into top of skull that allow measurement of electrical signals from 90 neurons (4 separate areas of her cerebral cortex)
    • In the long-term, HBMI’s will allow human brains to control artificial devices designed to restore lost sensory and motor functions ie do for the brain what the pacemaker did for the heart
    • Implants will help shed light on some of the brain’s mysteries: neuroscientists still know very little about how the electrical and chemical signals emitted by the brain’s millions of neurons let humans perceive color or smell, give rise to the precise movements of Brazilian soccer players
    • “We don’t have a finished model of how the brain words. All we have are first impressions”
    • Nicolelis’ latest experiments show that by tapping into multiple neurons in different parts of the brain, it is possible to glean enough info to get a general idea of what the brain is up to
      • For the monkey, it’s enough info to detect the monkey’s intention of making a specific movement a few tenths of a second before it actually happens
      • Nicolelis’ team succeeded at reliably measuring tens of neurons simultaneously over several moths –previously a key technological barrier –that enabled the remarkable demonstration with robot arm
    • Remaining challenges: developing electrode devices and surgical methods that will allow safe, long-term recording of neuronal activities
    • Nicolelis is working on developing a telemetry chip that would collect and transmit data through the skull, without unwieldy sockets and cables

2. Flexible Transistors

  • Implementations of pervasive computing will require integrated circuits that are both cheap and flexible (tough for today’s silicon technology”
  • Scientists working on transistors based on organic (carbon-based) molecules or polymers (organic electronics are inexpensive to manufacture and compatible with plastic substrates); however, organics are far slower than silicon cousins
  • Breakthrough: Cherie Kagan made a compromise: transistors made from materials that combine the charge-shuttling power and speed of inorganics with the affordability and flexibility of organics
  • Hybrids may be far faster than amorphous silicon, and have a key advantage over silicon-based electronics: it can be dissolved and printed onto paper or plastic like particles of ink
  • Kagan’s transistors could compete with organic electronics in variety of applications like radio-frequency product ID tags, flat-panel video displays (sharper images for a fraction of the cost) that lead to affordable wall-sized displays or high-quality displays that pop out of your pen (if all goes well, could be used in cheap, flexible displays within 5 years)
  • Bright displays that fit in your pocket will require portable power: Kagan’s newest interest: cheap, flexible materials for solar cells to liberate pervasive computing from bulky batteries

3. Data Mining

  • Data mining, also known as knowledge discovery in databases (KDD): a system that can burrow through gigabytes of website visitor logs in search of patterns no one can anticipate in advance to compile a simple recommendation list rather than sorting through a few megabytes of structured data to find answers to specific queries
  • Usama Fayyad, the pioneer behind data mining, was working at GM compiling a huge database on car repairs that would allow any GM service technician to ask the database  questions based on several car characteristics and get a response
    • Fayyad developed a pattern recognition algorithm to solve this, which was later used at NASA JPL to identify objects, and pursued by everyone from the military to doctors
  • Fayyad identified a need: companies needed someone to host their databases for them, and provide data-mining services on top, which led to him creating digiMine
  • Future: wide open as researchers move beyond original focus on highly structured, relational databases
    • Hot area: text data mining: extracting unexpected relationships from huge collections of free-form text documents.
      • UCB LINDI system has been used to help geneticists search biomedical literature and produce plausible hypotheses for the function of newly-discovered genes
    • Hot area: video mining: combining speech recognition, image understanding, and natural language processing to open up the world’s vast video archives to efficient computer searching
      • CMU Informedia II system is given CNN clips, it produces a computer searchable index by automatically dividing each clip into individual scenes accompanied by scripts and headlines

4. Digital Rights Management

  • Ranjit Singh, president of ContentGuard, spinoff of Xerox PARC, is on a mission to commercialize content protection in a wired world
  • Sits at ground zero of what may be bloodiest battle to shape the Internet during the 21st century’s initial decade: IP owners vs internet users (who want content to be freely distributed)
    • The internet allows perfect and totally frictionless distribution
  • Digital rights management (DRM) is the catalyst for a revolution in e-content, which will allow content owners to get much wider and deeper distribution than ever before, and see who is passing your content to whom
  • At its core, DRM amounts to an encryption scheme with a built-in e-business cash register, where content is encoded, and to get the key, a user needs to do something (ie pay money, provide an email address, etc). DRM providers deliver protection tools, whereas the content owners set the conditions
  • ContentGuard uses a multiple key approach; anyone receiving bootleg content would have to crack into all over again, so even if a hacker cracks a piece of content, he can’t distribute it
  • DRM isn’t ubiquitous for 2 reasons
    1. Content owners are in the midst of a hard rethink about both pricing and distribution: how do you price 3 listens to a song, or a download of a low-res image that can’t be forwarded to others? They are currently trying out different models for valuing content
    2. The user experience has to hide the complexity of protection technologies. Users have to be able to buy and consume content without jumping through hoops
  • Analysts don’t believe content can be protected in the Internet era; people want flexible access to content (re: Napster).
    • Napster is unstoppable, and even if courts stop it, the Internet’s enablement of frictionless distribution of digital content among millions will live on
  • The more content a business puts online, the faster it will want to put still more content up, because it will see the economic benefits and users will see the benefits of gaining access to more content, leading to a huge explosion (Network Effects)

5. Biometrics

  • Biometrics: identifying individuals by specific biological traits, has already emerged
    • Large companies use fingerprint sensors, facial recognition, iris-scanning
    • Consumers have been reluctant to adopt
  • Joseph Atick, President/CEO of Visionics (facial recognition), believes that the wireless Web will make consumer hungry for biometrics. PDAs and cell phones are becoming portals to users’ worlds, transaction devices, IDs, and maybe one day passports
    • With so much personal/financial information in one place, comes great need for security, which will drive biometrics. Security will drive need for biometric systems, other tech developments (increased bandwidth, camera phones, etc) will create infrastructure needed to put biometrics into consumer hands
    • Visionics is working to let people authenticate any transaction they make over the wireless Web using their own faces
  • Atick, while heading a lab at Rockefeller University, discovered that the brain deals with visual info much as computer algorithms compress files: because everyone has 2 eyes, a nose, lips, the brain extracts only those features that typically show deviations from the norm (ie bridge of nose, upper cheekbones), filling in the rest
  • Visionics develops FaceIt, which verifies a person’s ID based on a set of 14 facial features unique to an individual and unaffected by presence of/changes in facial hair/expression
    • Successfully used to fight crime in England and election fraud in Mexico
    • Signed merger agreement with Digital Biometrics to build the first line of “biometric network appliances” –computers hooked to the Net with capacity to store/search large databases of facial/biometric info. Appliances with customer ID data can receive queries from companies wanting to authenticate e-transactions. Accessing the system works with PDAs/desktops, but most will come from handheld devices
    • Also working with companies in Japan/Europe so consumers can capture their own faces and submit encrypted versions over the Net
  • Future: bringing back an old element of human commerce –restoring confidence that comes with doing business face to face
    • It will be 2-3 years before PDA and cell phone wielders will use biometrics instead of passwords and PINs

6. Natural Language Processing

  • New generation of interfaces arising that will allow extended conversation with computers; requires integration of speech recognition, natural language understanding, discourse analysis, world knowledge, reasoning ability, speech generation
    • DARPA working on interfaces that will ultimately include pointing, gesturing, and other forms of visual communication
    • IBM/Microsoft want a speech enabled “intelligent environment” where every object big enough to hold a chip actually has one; speech recognition necessary because they will each be too small to have a keyboard
    • Karen Jensen, chief of NLP at MSFT Research previously at IBM and contributed to MSFT’s Encarta encyclopedia and grammar checker, is now focused on MindNet, a system for automatically extracting a massively hyperlinked web of concepts from something like a standard dictionary
      • Let’s say a dictionary defines a motorist as “a person who drives a car”
      • MindNet uses automatic parsing tech to find definition’s underlying logical structure, identifying “motorist” as a kind of person, “drives” as a verb taking motorist as a subject and car as an object
      • Wants a conceptual networking tying together all of human understanding in words; show how 2 sentences said differently can mean the same thing
    • MindNet has proved to be great for translation –have 2 separate conceptual networks for English and a second language, then align the webs so English logical forms match other language equivalents –then annotate matched logical forms with data from English-other language translation so translation proceeds in either direction

7. Microphotonics

  • Photonic crystals are on the cutting edge of microphotonics: tech for directing light on a microscopic scale that will make a major impact on telecommunications
  • Goal: replace electronic switches with faster, miniature optical devices
    • None have the technical elegance and widespread utility of photonic crystals
  • Photonic crystals provide means to create optical circuits and other small, inexpensive, low-power devices that can carry, route, process data at the speed of light
    • Trend: make light do as many things as possible, won’t completely replace electronics though
  • Photonic crystals are to photons what semiconductors are to electrons: offering an excellent medium for controlling the flow of light
    • Crystals admit or reflect specific photons depending on wavelength and crystal design
  • MIT Prof John Joannopoulos suggested that defects in crystals’ regular structure could provide an effective and efficient method to trap the light or route it through the crystal
    • Mold the flow of light by confining light and figuring out different ways to make light bend, go straight, split, come back together in the smallest possible space
    • Breakthrough: Explained how crystal filters could pick out specific streams of light from the flood of beams in wavelength division multiplexing (WDM), tech used to increase amount of data carried per fiber
    • Helped set the stage for the world’s smallest laser and electromagnetic cavity, key components in building integrated optical circuits
  • Even with an all-optical Internet, other problems loom:
    • Advancements due to improving fibers and tricks like WDM, but in 5-10 years, experts fear it won’t be possible to squeeze any more data into fiber optics
      • “Perfect mirror” photonic crystals may be the solution: reflect specific wavelengths of light from every angle with extraordinary efficiency. Hollow fibers with this reflector could carry up to 1000x more data than current fiber optics. It doesn’t absorb/scatter light like glass, so it could also eliminate expensive signal amplifiers needed every 60-80km for today’s optical networks
  • What are the theoretical limits of photonic crystals?? How much smaller can they be made? How can they be integrated into optical chips for telecom/computers?
  • Once you start being able to play with light, a whole new world opens up

8. Untangling Code

  • Gregor Kiczales, scientist at Xerox PARC, champions “aspect-oriented programming,” a technique that will allow software writers to debug code as easily as it is used by laymen
    • “Crosscutting” refers to capabilities, like logging and security and synchronization, that are the same kind of shortcuts those in other professions have been using for a while
    • Logging: ability to trace and record every operation an application performs; only works if people remember to follow it
    • Security and Synchronization: ability to make sure that 2 users don’t try to access the same data at the same time; requires programmers to write the same functionality into many different areas of the application
    • Keeping track of crosscutting concerns is error-prone; forget to upgrade just a few instances, and bugs start to pile up
  • Kiczales proposes a new category in a language called an “aspect,” which allows programmers to write, view, and edit a crosscutting concern as a separate entity
    • Meaning less buggy upgrades, shorter product cycles, better/cheaper software
  • Many firms already have a version, but Kiczales is the first to take it to the real world by incorporating it into a new extension of Java
    • Northeastern: adaptive programming; IBM: subjective programming; University of Twente: composition filtering; Elsewhere: multidimensional separation of concerns
  • Major changes in programming methodology can take 30 years for acceptance, aspect could cut that cycle down by 15-20 years.
  • Crosscutting concerns aren’t actually hard to work with, once you have the proper programming support

9. Robot Design

  • Big obstacle: expensive to design and make robots smart enough to adapt readily to different tasks and physical environments the way human being do. How do builders build more complexity into robots without custom-tailoring each one?
    • Robots stuck in commercial niche doing simple, repetitive jobs (ie assembly line, mass production of toys, etc)
  • Promising approach: automate the design and manufacture of robotics by deploying computers to conceive, test, and even build the configurations of each robotic system
    • Jordan Pollack of Brandeis, directed a computer to design a moving creature using a limited set of simple plastic parts: plastic rods, ball joints, small motors, and a “brain” (neural network)
      • The computer –using an algorithm inspired by biological evolution –evolved hundreds of generations of potential designs, killing off the sluggish and refining the strong; bringing to life the strongest with a rapid-prototyping machine
      • Important point of coevolutionary design and automated manufacturing for robotics is to get small-quantity production to be economical (expects first cheap industrial robots to be 5-10 years away)
  • Before robots reach out into everyday world of businesses and households, they need their own version of Moore’s Law: becoming dramatically more affordable and powerful over time
    • Designing even relatively simple robots is a painstaking task: Honda has spent 14+ years building a humanoid robot able to walk, open doors, navigate stairs

10. Microfluidics

  • Microfluidics: a promising new branch of biotechnology with the idea that once you master fluids at the microscale, you can automate key experiments for genomics and pharmaceutical development, perform instant diagnosis tests, even build implantable drug-delivery devices –all on mass-produced chips
    • Microfluidics will do for biotech what the transistor did for electronics
  • Problems: developing general tech that can be used for a broad range of applications with several functions to be integrated into a single chip. Manufacturing, particularly silicon micromanaging, is so expensive that experts question if the products using these techniques can ever be economical to manufacture
    • Stephen Quake’s group (at CalTech), unveiled a set of microfabricated valves and pumps –a critical first step in developing tech general enough to work for any microfluidics application
      • To make microfluidics cheaper, Quake is casting them out of soft silicone rubber in reusable molds (“soft lithography”)
      • Potential for mass-produced, disposable microfluidic chips that make possible everything from drug discovery on a massive scale to at-home tests for common infections
      • First was a microscale DNA analyzer that operates faster and on different principles than conventional full-sized version, then a miniature cell sorter, and most recently, those valves and pumps
  • Quake finished bachelors and masters at Stanford in physics in 4 years, got bored, and started focusing on “the physics of biology” and was hired at CalTech as the first interdisciplinary Professor, and gained tenure at just 31 years old
  • Quake founded a startup called Mycometrix, which has licensed all of Quakes microfluidics patents from Caltech, and is planning to deliver its first microfluidic devices to researchers soon (HP and Motorola are trying as well, but only Mycometrix has actually brought a product to market)
  • Quake more interested in basic biology questions: How do the proteins that control gene expression work? How can you do studies that cut across the entire genome?
    • Now that Quake has some neat tools, he’s looking to do some science with them
  • Quake is the prototypical innovator: he has ability to work in all areas, from basic research to hot commercial markets.

Some Additional Thoughts:

  • It’s been almost 19 years since this publication came out: it’s interesting to see that many of the scientists predictions in terms of time have been off. It seems like they ran into science vs engineering problems (separate post on this later).
    • Researchers back then said breakthroughs were 5-10 years away. One has to wonder which breakthroughs scientists are saying are right around the corner are actually right around the corner, and which are still far off dreams (re:autonomous driving).
  • Many of the technologies highlighted are being highlighted today; Not sure if they went through a winter and are currently going through a resurgence, or if they have been hot his entire time. Are technology hype cycles 10 years? 20 years?
  • Good mix of university researchers and startups. Today hot startups/companies working on these technologies include:
    • Brain-Machine Interfaces: Neuralink, CTRL-Labs
    • Biometrics: Amazon, Apple, Face++, SenseTime, Alibaba
    • Natural Language Processing: IBM, Bytedance, Facebook
  • Do researchers make good entrepreneurs? At the very least, most of the startups mentioned here seem to be out of business or now part of other companies. But is this due to natural life cycle of tech companies or researcher competency?
  • What kind of breakthroughs need to happen to actually make technologies reality? We know neural networks in their current form (backwards propogation) have been around since the 1980s. Advanced sensors and chips are what led to the massive collection of data and explosion in computing power that has driven the AI boom of the last decade (Thanks Nvidia, Thanks Google).
    • What are the key drivers of a technology?
    • When will the catalyst(s) that propel them forward occur?
    • Who is leading the charge for each field?
    • Why do these technologies even matter? Do they even matter?
    • Where will these breakthroughs happen? Where geographically? Which disciplines? Cross border? Interdisciplinary?
    • How will these innovations change the way we live? Will they help us thrive, or just survive?

Stanley Druckenmiller: Generational Theft

On 5/7/2013, Stanley Druckenmiller and Geoff Canada gave a presentation at Bowdoin College titled: Generational Theft. You can watch the entire thing here. These are my notes. Additional thoughts/takeaways at the end.

  • Entitlements:
    • Websters: A right you have under the Law
    • Actual: A benefit that people have under the Law that they don’t have to provide a current service for
      • Income Supplement (lots of Press, but not really anything)
        • Unemployment insurance claims
        • Food stamps
      • Three Main Buckets (big big money) – Primarily for the Elderly
        • Medicaid: Income based (70% goes to elderly)
        • Social Security
        • Medicare
  • Druckenmiller started worrying about entitlements in 1994
    • In 2011, the baby boomers, the front end of what was going to turn 65, and there would be a huge surge in entitlement payments
  • Whats the demographic issue?
  • Chart 1: Federal Government Entitlement Transfers as a Percentage of Federal Budget Outlays
    • Services/Benefits people get while not providing a current service
    • 1960: 28% of Federal Outlays
    • 2010: 68% of Federal Outlays
    • Used to be 1/3 of Defense, now much more
    • Biggest increase took place in Nixon, Ford, GWB admins (NOT a partisan increase)
    • Elderly taking a bigger and bigger share even BEFORE baby boomers reach that age

  • Chart 2: Benefit per oldster/GDP per capita
    • Benefit/Oldster has grown MUCH quicker than GDP per capita since the 70s
      • 1970: 41%
      • 2010: 72%
  • Chart 3: Average Consumption by Age Group
    • People receiving outlays are about to explode
    • 1960s: 20s consumed much more than people in their 70s
    • 1990s: 70s consumer much much more than people in the 90s
    • AARP taking money from you and giving to the elderly
  • Chart 4: US Poverty Rates by Age Group
    • Elderly: 35% -> 9%
    • Children: 25%  -> 21%
    • Minority Children: 31% -> 35%
    • Massive wealth transfer from young to elderly
  • Chart 5: Average Children per woman
    • Fertility Rates:
      • 1950s: 3.7
      • 2011: 2.06
    • In 1957, there were 100MM less people in the US, and they were having more babies than we are today
      • Those babies are about to become seniors, who have been taking more and more, and there are a lot of these babies about to become seniors
  • Chart 6: US Life Expectancy at Birth (Years)
    • 1900 to Now
      • Men: 47.9 -> 74.9
      • Women: 50.7 -> 79.9
    • Not only will there be MORE oldsters, they will be around for a much longer period of time
      • Seniors:
        • A) Bigger Share
        • B) More of them
        • C) Taking for a longer period of time
  • Chart 7: Federal Spending by Entitlement’s Program (% of GDP)
    • Percentage of GDP
      • 3 Buckets: MMSS
        • Now to 2050: 10% to >20%
    • Historically, tax revenues as % of GDP have a ceiling around 20%
      • Entire tax revenues will be going to people who don’t even provide a service
  • Chart 8: US Population Ratio: Working Age to 65 and Over
    • By 2030, there will be HALF the amount of young to support the old
    • Working age: 18-64: Grow 17% from 2010-2050
    • Elders: 65+: Grow 102% from 2010-2050
    • 4.8 workers supporting elders
      • 2030: 2.9
      • 2050: 2.4
    • Huge recipient pool, less people giving money to
  • US Centenarians (Thousands) are the fastest growing age population
    • Oldsters growing at 102%
      • 85+ growing at 343%
      • 100+ growing at >1000%
        • Spend A LOT more on medicare than people in 60s and 70s
  • Chart 9: The True “Fiscal Gap”
    • Total Debt (On the Books): 11Trillion
    • Fiscal Gap: 211Trillion
      • Expected tax revenues – present value of benefits promised
    • Labor: Pay a payroll tax: supposedly to pay social security and future medicare: but it doesn’t pay for you, it pays for the current elders
      • All you have is a promise you’ll have some
    • Deficit: 10% of GDP, actually 15% of GDP
  • Chart 10: Federal Taxes/Expenditures
    • The increase in federal taxes and expenditures needed to cover fiscal gap today is much lower than in 40 years
    • Increase in all federal taxes:
      • Today: 64%
      • 20 years: 77%
      • 40 years: 93%
    • Decrease in all federal expenditures (defense, entitlements, etc)
      • Today: 40%
      • 20 years: 46%
      • 40 years: 53%
  • Chart 11: Raising taxes on the rich is not the solution
    • 10%, 1000bp Rise is Millionaire Tax Rate: $92B
    • 2010 Deficit: 1.3Trillion (not including off balance sheet)
    • At 50%, every day you work until June 30th, all your income goes to someone else
      • You don’t get anything till July 1st
    • If you tax the rich, they:
      • Stop working
      • Move
    • Even if you get the money, it doesn’t move the needle
  • Chart 12: 2011 US Defense spending exceeded the 13 next highest defense budgets combined
    • USA: 711B
    • China: 130B
    • Entitlements growing MORE than 711B, so if you cut 100%, still screwed
  • Chart 13: Net worth by age of household head in 2007
    • 20-34: 100K
    • 65-74: 1MM
  • Chart 14: US Consumption by age (ratio to labor income ages 30-49)
    • 1960:
      • X-axis: Age
      • Y-axis: What the average 30-49yr old makes in
    • 90yr old is spending 138% what 40 year olds makes
      • Spending 2x what 30 year olds spend
      • Lots to keep elders alive an extra 5 days
  • Chart 15: National Health Spending across countries
    • AN ADDITIONAL PROBLEM OF INEFFICIENCY. US not even close to any other country
    • US: 17.6
    • Average: 9.5
  • Chart 16: Despite spending 2x average country, health outcomes are bad
    • Spending the most, yet near the bottom for health outcomes
    • Why? Screwed up system: Malpractice insurance scam of an industry
      • Lawyers taking a bigger bite from physician income – NOT the biggest problems
      • Incentivizes hospitals to run insane (10-15 tests) that studies have shown don’t actually make an impact on end result
        • CYA policy (cover your ass)
      • Medical bills all done through insurance – people have no idea what they pay
        • HAVE TO force people to shop (aka know the prices), otherwise things won’t improve
  • Chart 17: US National Saving and Domestic Investment Rates (%)

Some Additional Thoughts:

  • As of this post, it has been 6 years since Druckenmiller went on his college campus tour. Unfortunately, it appears that he was unable to inspire the youth to go out and fight for their future.
  • Interestingly, the Medicare and Medicaid have been core issues that politicians have been campaigning around. However, there has yet to be a convincing payment proposal. The wealth tax in particular has been a central focus for the Democratic Party in the run-up to the 2020 election.
  • It is increasingly clear that the ElderTech market will be far bigger than anyone imagined. However, there has yet to be a majorly-funded ElderTech startup, nor the establishment of any ElderTech venture capital funds/teams.
  • Historically, America has hidden (or solved, depending on who you ask) its demographic issues by taking in immigrants. However, since the 2016 US Presidential Election, relying on immigrants is becoming less and less likely, whether that be for blue collar jobs or white collar jobs.
  • America’s demographics seem to be following in the footsteps of Japan, Korea, the Scandinavian countries, and many other countries: less babies, less immigration. Will this lead to a stagnating US economy?
  • Given the increasing Elder population and shrinking Youth population, it is more important now than ever to invest in our children. Currently seeking out children/education-focused startups.
  • Note: Thought of the Malthusian Trap and a running out of resources. Similar to the industrial revolution, the implementation of AI will likely increase human productivity by leaps and bounds. But will it be enough?