SITALWeek #197
Stuff I thought about last week 6-16-19
Greetings – Are tech companies responsible for how their algorithms are used? Facebook has been saying this problem is the responsibility of global governments, and this week Amazon joined that chorus in denying responsibility, calling on the government amidst a slew of unsettling facial recognition news; the failure of the underpinnings of Modern Portfolio Theory – still used by many active managers – is evident in new research on ergodicity; the problem with ceding art to AI; and a slew of cool robot, drone, and tech news. As always, reply back with thoughts or grab me on Twitter.
Click here to SIGN UP for SITALWeek’s Sunday EMAIL (please note some ad blocking software may disrupt the signup form; if you have any issues or questions please email sitalweek@nzscapital.com )
Stuff about Innovation and Technology
Infarm is a “farming as a service” company that uses the sensors and the cloud to run vertical, indoor farms in grocery stores, schools and other locations. The modular “farms” eliminate some of the environmental footprint and improve on the 45% of nutrients lost in transporting fresh food. So far, the German company has farms in 350 locations harvesting 150,000 plants monthly.
Chinese drone king DJI launches a new robot tank, and the video is pretty cool. Of course, if we are entering a technological war with China, what better way for China to infiltrate the US than fill every household that has kids with pellet-firing, laser-wielding autonomous tanks!
I recently discussed how some McDonald’s are seeing 15% of their business from Uber Eats deliveries, and soon those Big Macs might be arriving by Uber drones. Until the technology is more mature (and safer!), Uber will use drones to position food in safe “landing zones” where couriers will then run them shorter distances to hungry folks.
Decreasing the costs of robotics is critical to their adoption. London-based Automata’s new Eva robotic arm has 80% of the precision of Japanese robots for 20% of the cost. It’s also easy to program in just minutes from any device.
Tesla is finally getting close to launching their solar roof tiles. The goal is to ultimately get solar roofs at parity with regular roofs. In the meantime, Tesla is focused on aesthetics: the company will not only install the new solar tiles but replace vents, skylights, and flashing to match.
This week in “facial recognition anxiety” news: 1) China’s largest insurance company Ping An Insurance uses facial recognition of microexpressions to determine truthfulness of customers and employees; 2) Vuzix introduces autonomous facial recognition glasses currently in use by security forces in the UAE; 3) the US Customs and Border Protection announced a breach of 100,000 images of travelers crossing the border by car; 4) how to opt out of facial recognition at airports; 5) and I declined to have my daughter’s photo captured in order to be used for faster access to a popular theme park...
However, by far the most disturbing comments on facial recognition came from Amazon chief technology officer Werner Vogels in this BBC article:
Mr Vogels doesn’t feel it's Amazon’s responsibility to make sure Rekognition is used accurately or ethically.
“That’s not my decision to make,” he tells me.
“This technology is being used for good in many places. It’s in society’s direction to actually decide which technology is applicable under which conditions.
“It’s a societal discourse and decision - and policy-making - that needs to happen to decide where you can apply technologies.”
He likens ML and AI to steel mills. Sometimes steel is used to make incubators for babies, he says, but sometimes steel is used to make guns.
So, if a hostile government uses Amazon’s Rekognition tool to commit the genocide of a million humans, I guess Werner will still sleep well? A better approach would have Amazon requiring an application and attestation of the technology’s use, especially until it’s ready to be deployed in real-world situations that impact lives. How hard would it be for Amazon to have an ethics committee determining fair use policies that said, for example, using Rekognition for finding victims of sex trafficking is ok, but using it to put someone in jail before the technology is fully ready (which it's not) is not ok? Werner’s steel analogy is a very flimsy straw man: when raw steel leaves a factory, it’s no longer traceable, but every AWS algorithm is running on Amazon servers at all times, making it easy to track and regulate. I don't buy this lame excuse from companies like Facebook and Amazon that only the government has responsibility over how their products are used. If that's really true, then Zuckerberg and Bezos should sign their shares and voting rights over to the government.
Assistant Attorney General in the Department of Justice Makan Delrahim gave a speech last week on antitrust and the big tech platforms (a topic we covered in last week’s SITALWeek and in more depth here). It’s interesting to read the entire speech, but one thing really stuck out to me: a near-complete lack of discussion on data and privacy. There was one comment – that increased competition could mean increased protection of data privacy – but I didn’t feel the concept was fleshed out in the speech. The analysis ignored the fact that data have real value to consumers when used intelligently. I don’t want two social networks competing to use less of my data, I want a service that gives me control and allows that data to be used how I want for my benefit. The discussion and reference to network effects struck me as lacking as well. Overall, this speech seemed to ignore the important differences between information- vs. asset-based businesses; it felt like a legacy, Industrial-Age view of competition, which I believe to be a very dangerous lens to analyze data-driven, network-effect platforms.
Last week I mentioned the progress China is making on cancer drugs; this week I read that the FBI and NIH are purging Chinese cancer researchers from universities and institutions in the US. Medical research is increasingly a data-driven, network-effect science, and cutting off global collaboration and sources of data may have very negative consequences to progress.
The average cost of a server has doubled over the last decade to nearly $8,000. Increasingly rich configurations with more memory, GPUs, and FPGAs along with x86 advancements have no doubt increased the performance of servers, but the price/performance has been relatively flat for the last few years. The magnitude of the drop in server demand from Q4 2018 to Q1 2019 was the largest seen in a decade. Last week at the Merrill Lynch Global Tech Conference, the head of high-performance computing for AI-chip maker NVIDIA had this to say regarding the slowdown in data center spending at the hyperscale cloud platforms: “[the slowdown] means their AI teams have to continue to use the infrastructure that they have. Demand doesn't slow down, so they get angrier. These are some of the most important people inside of those companies. And so they're going to naturally want more of their AI infrastructure back...So we fully expect that to just be cyclical and come back in the latter half of the year.”
I mentioned French semiconductor material science company Soitec last week, and this week the company held an extensive capital markets day for investors. If you’re interested in the cool materials enabling chips for the IoT, data centers, EVs, etc., then you might want to skim their presentation deck (PDF). The Soitec substrates help circumvent many power constraints that traditional silicon bumps up against.
Can analog semiconductors compute AI workloads better than big digital chips? There are several startups in this area, but it’s a challenge to put tens of thousands of digital-to-analog converters (DAC) and analog-to-digital converters (ADC) on a single chip, as compared to the typical 8-16 in an analog device. When the workload is in between the DAC and ADC, the analog circuitry can theoretically inference an answer vs. a machine-learning model at lower power. Also, this chip can be manufactured on older 40nm technology (as opposed to the 7nm seen in leading-edge GPUs and FPGAs).
Spotify uses your mood, based on the music you are listening to, for data sharing and ad targeting.
In other music news, Venture capitalist Vinod Khosla suggested that in 10 years all music will be custom made for individuals by AI. That’s a world I’m not interested in – in my view, AI should be focused on the things humans aren’t good at, and humans should be focused on more creative endeavors. It reminds me of the quote from storyteller and magician Penn Jillette: "The fact of the matter is, since the beginning of time, you could buy a Picasso and change the colors. That's trivial. But you don't because you're buying a piece of Picasso's fucking soul. That's the definition of art: Art is one person's ego trip." I’d suggest it’s this connection to an artist’s unique “soul” that makes music and other mediums resonate, so I would let AI handle the technology, and leave the composition to the humans.
Speaking of art, BroadwayHD is a company trying to stream Broadway shows to viewers around the world, and it’s created a lot of debate in the theater world. That debate strikes me as fairly myopic; I think opening up a medium to new consumers can only grow the potential market. Certainly, live events are increasingly valuable and important; however, access to streaming seems much more likely to grow future live audiences than shrink them.
Simulated violence in video games, I believe, is an artistic portrayal that does not have negative consequences to players (and research over the last 25 years seems to support that view). However, perhaps the boundaries are being pushed a little too far: this article on the new Call of Duty: Modern Warfare is a little hard to stomach. On the other hand, maybe, as the developers suggest, players will have more empathy for soldiers with PTSD (and other victims of warfare) and develop a profound, visceral aversion to violence.
Here is a great profile on Tim Sweeney, founder of Epic Games, which created the hit, industry-changing Fortnite. And here is another game industry exec profile on Tucker Roberts, the son of Comcast’s CEO and head of the company’s esports division, which has some interesting insights into the world of professional esports for those interested in that rapidly growing – but near-term overhyped – industry.
Internet censorship in China was in full force this week, as Hong Kong citizens fought to maintain their right to not be capriciously extradited to mainland China. Residents of mainland China were largely left in the dark; in one particularly funny story, the state’s main paper reported 800,000 people marched in favor of extradition (in reality, more than a million HK citizens continue to take to the streets to protest the proposed law, which for now has been postponed). For candidates running in Taiwan’s upcoming presidential election under the banner of closer ties to China, these protests might be a precursor to subsequent civil unrest on the island of Taiwan as well. Related: a chilling look at the far reaches of Tencent’s Wechat app.
Big money comes to house flipping: in another example of low rates attracting yield-seeking investors to riskier asset classes, KKR invests $250M into a company with $1.5B in loans out to house flippers. I’ve been involved in this market for a couple of years through PeerStreet, and the high-single-digit rates available in this market underline just how risky this type of investing is in our very low global rate environment, especially ten years into a US housing market recovery.
Miscellaneous Stuff
Nick Hanauer (who frequently collaborates with one of our favorite economists Eric Beinhocker) argues that “income inequality has exploded not because of our country’s educational failings but despite its educational progress”. To reverse the trend of economic equality, he says: “...most important, if we want to build the sort of prosperous middle-class communities in which great public schools have always thrived, we must pay all our workers, not just software engineers and financiers, a dignified middle-class wage.” This is a powerful inversion in the way the West should be thinking about inequality.
Norway will sell $7.5B of oil stocks but still maintain large positions in the big oil companies as the country continues – with a little touch of hypocrisy – its commitment to the environment. The $1T sovereign wealth fund, built up in no small part on the sale of oil, has been unable to get full government support to divest oil stocks. This despite electric vehicles accounting for more than half of new passenger car sales this year in the country, and 95% of electricity in Norway being generated by hydropower.
A somewhat startling stat about the environmental cost of travel: “Each additional metric ton of carbon dioxide or its equivalent — your share of the emissions on a cross-country flight one-way from New York to Los Angeles — shrinks the summer sea ice cover by 3 square meters, or 32 square feet.”
Trying to recreate the distinct Lego “click” and pull-apart in non-plastic materials is proving to be very difficult. The Lego Group is investing hundreds of millions of dollars, but they haven’t yet had a breakthrough. In related plastic news, Canada will ban single-use plastics they deem harmful (bags, straws, utensils, stir sticks, etc.) by 2021.
The CEO of Ball Corp made the following comments – regarding the company’s forthcoming aluminum cup – earlier this month at the Deutsche Bank Global Industrials & Materials Summit:
“This is [an] infinitely recyclable aluminum cup made out of the same material that this [soda can] is made of...we're able to make this at a price point that's very competitive with compostable plastic. And the irony of our compostable plastic is only a portion of that is compostable, so I don't know what the other half happens, other than polluting our oceans and worlds, but this is completely recyclable. And when we think this has -- just give you a sense, we estimated there is not great data on it, but it's probably about a 60-plus-billion unit in North America alone. We're starting out with a pilot. We're launching the pilot with a couple of institutions near our home in Colorado. And then as part of that, there is sports venues and other things, and we said we're delighted to have you start with us on one condition that you go plastic-free. And so when you think about sports venues, when you think about college campuses, that's kind of our initial target, but we're also looking much broader. You can go and buy plastic, 50-sleeve of plastic cups for $12 on Amazon last time I checked. We can sell at the same price this type of thing that's so much more from a sustainability perspective, so much better. And so we're on the very front edge of this. We're actually designing a ramp-up of full commercial capability. We're probably 12 months away from that, but we have high hopes and high expectations.”
Stuff about Geopolitics, Economics, and the Finance Industry
Expected Utility Theory (EUT) and Prospect Theory from economics and behavioral psychology, respectively, are not useful when it comes to actual human decision making, yet both continue to be used by investors either intuitively or explicitly. Specifically, Modern Portfolio Theory (MPT), developed by Markowitz in 1952 and used by many portfolio managers today, is an extension of the heavily-flawed EUT. These theories are flawed largely because they assume ergodic behavior. Ergodic systems have the same expected value as the average value: if you toss a coin a hundred times, you expect 50/50 heads and tails, which is the same as the expected probability of any given toss. Non-ergodic systems, in contrast, take into account the path through time, not just the probabilistic outcome. For example, you could potentially get 20 tails in a row (albeit at a low probability), and, if you were betting on those tosses, the multiplicative damage that might do to your wallet if you are doubling down on every bet is potentially catastrophic (as Mr. Howell discovered when Gilligan kept sinking those improbable putts in the episode “Three Million Dollars, More or Less”!). This new paper (PDF) shows that humans are much more risk averse in these multiplicative outcomes, which seems obvious, yet it’s nowhere in traditional economic theory and often missing from discussions of cognitive bias. Accounting for non-ergodicity is a key underpinning of our Resilience+Optionality portfolio construction process, and we discuss it on page 8 of our Complexity Investing whitepaper.
S&P has removed Facebook from their "ESG" index. Owning Facebook as an ESG investment to begin with was wronger than Wrongway Feldman. Let's start with “G” and work our way backwards: Facebook has zero governance and zero accountability to its fiduciary duties as a result of Zuckerberg's voting control. "S"? Let's not even start with the negative social impacts the company continues to have and Zuckerberg's insistence that it's the government's fault, not his. And "E"? Who cares at this point, 2 strikes already. None of this is new, so herein lies the problem with taking a passive approach to socially responsible investing: you'd be overweight Facebook compared to the regular S&P500. How many other mistakes has the S&P made in constructing the ESG index? “To live outside the law you must be honest”: the law has a narrow and anachronistic definition of fiduciary duty, so, if you want to be honest, do what's right: dig in and find the companies really moving society forward, don’t just rely on what some agency claims to qualify for social responsible investing.
Fund managers would be much better off selling stocks at random, according to this paper from earlier this year, which found that value is created buying stocks, whereas most selling strategies tend to destroy value. This would also seem to support the winning strategy of low turnover and long-term holding – the longer the hold, the fewer times you have to make a sell decision!
The rise of actively managed ETFs, which print their portfolios daily, forces an interesting philosophical question for portfolio managers: what percentage of your value to investors is from portfolio construction and stock selection versus what percentage is from your ability to overcome the cognitive bias and psychological mistakes most investors make when timing the buying and selling of stocks? I think most of us would have said it’s 80% stocks and portfolio combined with 20% instinct and bias-avoidance (e.g., being greedy when others are fearful, avoiding anchoring, etc.). But, perhaps it’s the other way around: maybe picking stocks and assembling a risk-bearing portfolio is the easier part, and the real value that merits an active management fee is the psychology to successfully navigate the market’s ups and downs? If I can see a multitude of well-constructed portfolios with thoughtful stock selection daily in active ETFs, then really what investors are now paying for is manager experience and fortitude to exploit market volatility without making common mistakes of cognitive bias. However, we are on the cusp of the rapid rise of opaque, active ETFs that will be able to hide their portfolios. The broader trend in the global economy is rising transparency and power of information. The opaque ETF runs counter to this, and I’d suggest that, although it’s an intriguing concept (and should replace all mutual fund structures due to its tax advantages), it seems to belie the truth that the majority of value add likely comes from experience with bias and decision making and not secret portfolios. I can print a portfolio every day, but will people have the fortitude to copy it, or will they second guess every change they see?
Disclaimers:
The content of this newsletter is my personal opinion as of the date published and are subject to change without notice and may not reflect the opinion of NZS Capital, LLC (“NZS”). This newsletter is simply an informal gathering of topics I’ve recently read and thought about. It generally covers topics related to the digitization of the global economy, technology and innovation, macro and geopolitics, as well as scientific progress, especially in the fields of cosmology and the brain. I will frequently state things in the newsletter that contradict my own views in order to be provocative. I often I try to make jokes, and they aren’t very funny – sorry.
I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC (“NZS”). If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital has no control. In no event will NZS be responsible for any information or content within the linked sites or your use of the linked sites.
Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results.
Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.