We’ve constantly been sharing a list of our recent reads in our weekly emails for The Good Investors.
Do subscribe for our weekly updates through the orange box in the blog (it’s on the side if you’re using a computer, and all the way at the bottom if you’re using mobile) – it’s free!
But since our readership-audience for The Good Investors is wider than our subscriber base, we think sharing the reading list regularly on the blog itself can benefit even more people. The articles we share touch on a wide range of topics, including investing, business, and the world in general.
Here are the articles for the week ending 06 August 2023:
1. The Next Frontier For Large Language Models Is Biology – Rob Toews
One striking theme has emerged from the long march of research progress across biochemistry, molecular biology and genetics over the past century: it turns out that biology is a decipherable, programmable, in some ways even digital system.
DNA encodes the complete genetic instructions for every living organism on earth using just four variables—A (adenine), C (cytosine), G (guanine) and T (thymine). Compare this to modern computing systems, which use two variables—0 and 1—to encode all the world’s digital electronic information. One system is binary and the other is quaternary, but the two have a surprising amount of conceptual overlap; both systems can properly be thought of as digital.
To take another example, every protein in every living being consists of and is defined by a one-dimensional string of amino acids linked together in a particular order. Proteins range from a few dozen to several thousand amino acids in length, with 20 different amino acids to choose from.
This, too, represents an eminently computable system, one that language models are well-suited to learn.
As DeepMind CEO/cofounder Demis Hassabis put it: “At its most fundamental level, I think biology can be thought of as an information processing system, albeit an extraordinarily complex and dynamic one. Just as mathematics turned out to be the right description language for physics, biology may turn out to be the perfect type of regime for the application of AI.”…
…Proteins are involved in virtually every important activity that happens inside every living thing: digesting food, contracting muscles, moving oxygen throughout the body, attacking foreign viruses. Your hormones are made out of proteins; so is your hair…
…As mentioned above, every protein consists of a string of building blocks known as amino acids strung together in a particular order. Based on this one-dimensional amino acid sequence, proteins fold into complex three-dimensional shapes that enable them to carry out their biological functions.
A protein’s shape relates closely to its function. To take one example, antibody proteins fold into shapes that enable them to precisely identify and target foreign bodies, like a key fitting into a lock. As another example, enzymes—proteins that speed up biochemical reactions—are specifically shaped to bind with particular molecules and thus catalyze particular reactions. Understanding the shapes that proteins fold into is thus essential to understanding how organisms function, and ultimately how life itself works.
Determining a protein’s three-dimensional structure based solely on its one-dimensional amino acid sequence has stood as a grand challenge in the field of biology for over half a century. Referred to as the “protein folding problem,” it has stumped generations of scientists. One commentator in 2007 described the protein folding problem as “one of the most important yet unsolved issues of modern science.”
In late 2020, in a watershed moment in both biology and computing, an AI system called AlphaFold produced a solution to the protein folding problem. Built by Alphabet’s DeepMind, AlphaFold correctly predicted proteins’ three-dimensional shapes to within the width of about one atom, far outperforming any other method that humans had ever devised.
It is hard to overstate AlphaFold’s significance. Long-time protein folding expert John Moult summed it up well: “This is the first time a serious scientific problem has been solved by AI.”…
…AlphaFold was not built using large language models. It relies on an older bioinformatics construct called multiple sequence alignment (MSA), in which a protein’s sequence is compared to evolutionarily similar proteins in order to deduce its structure.
MSA can be powerful, as AlphaFold made clear, but it has limitations.
For one, it is slow and compute-intensive because it needs to reference many different protein sequences in order to determine any one protein’s structure. More importantly, because MSA requires the existence of numerous evolutionarily and structurally similar proteins in order to reason about a new protein sequence, it is of limited use for so-called “orphan proteins”—proteins with few or no close analogues. Such orphan proteins represent roughly 20% of all known protein sequences.
Recently, researchers have begun to explore an intriguing alternative approach: using large language models, rather than multiple sequence alignment, to predict protein structures.
“Protein language models”—LLMs trained not on English words but rather on protein sequences—have demonstrated an astonishing ability to intuit the complex patterns and interrelationships between protein sequence, structure and function: say, how changing certain amino acids in certain parts of a protein’s sequence will affect the shape that the protein folds into. Protein language models are able to, if you will, learn the grammar or linguistics of proteins.
The idea of a protein language model dates back to the 2019 UniRep work out of George Church’s lab at Harvard (though UniRep used LSTMs rather than today’s state-of-the-art transformer models).
In late 2022, Meta debuted ESM-2 and ESMFold, one of the largest and most sophisticated protein language models published to date, weighing in at 15 billion parameters. (ESM-2 is the LLM itself; ESMFold is its associated structure prediction tool.)
ESM-2/ESMFold is about as accurate as AlphaFold at predicting proteins’ three-dimensional structures. But unlike AlphaFold, it is able to generate a structure based on a single protein sequence, without requiring any structural information as input. As a result, it is up to 60 times faster than AlphaFold. When researchers are looking to screen millions of protein sequences at once in a protein engineering workflow, this speed advantage makes a huge difference. ESMFold can also produce more accurate structure predictions than AlphaFold for orphan proteins that lack evolutionarily similar analogues.
Language models’ ability to develop a generalized understanding of the “latent space” of proteins opens up exciting possibilities in protein science.
But an even more powerful conceptual advance has taken place in the years since AlphaFold.
In short, these protein models can be inverted: rather than predicting a protein’s structure based on its sequence, models like ESM-2 can be reversed and used to generate totally novel protein sequences that do not exist in nature based on desired properties.
All the proteins that exist in the world today represent but an infinitesimally tiny fraction of all the proteins that could theoretically exist. Herein lies the opportunity.
To give some rough numbers: the total set of proteins that exist in the human body—the so-called “human proteome”—is estimated to number somewhere between 80,000 and 400,000 proteins. Meanwhile, the number of proteins that could theoretically exist is in the neighborhood of 10^1,300—an unfathomably large number, many times greater than the number of atoms in the universe. (To be clear, not all of these 10^1,300 possible amino acid combinations would result in biologically viable proteins. Far from it. But some subset would.)..
…Using AI, we can for the first time systematically and comprehensively explore the vast uncharted realms of protein space in order to design proteins unlike anything that has ever existed in nature, purpose-built for our medical and commercial needs.
We will be able to design new protein therapeutics to address the full gamut of human illness—from cancer to autoimmune diseases, from diabetes to neurodegenerative disorders. Looking beyond medicine, we will be able to create new classes of proteins with transformative applications in agriculture, industrials, materials science, environmental remediation and beyond…
…Thanks to scientific breakthroughs that have made gene sequencing vastly cheaper and more accessible over the past two decades, the amount of DNA and thus protein sequence data available to train AI models is growing exponentially, far outpacing protein structure data.
Protein sequence data can be tokenized and for all intents and purposes treated as textual data; after all, it consists of linear strings of amino acids in a certain order, like words in a sentence. Large language models can be trained solely on protein sequences to develop a nuanced understanding of protein structure and biology.
This domain is thus ripe for massive scaling efforts powered by LLMs—efforts that may result in astonishing emergent insights and capabilities in protein science.
2. Country Risk: A July 2023 Update – Aswath Damodaran
What makes some countries riskier than others to operate a business in? The answer is complicated, because everything has an effect on risk, starting with the political governance system (democracy, dictatorship or something in between), the extent of corruption in the system, the legal system (and its protection for property rights) and the presence or absence of violence in the country (from wars within or without)…
…Things get even more complicated when you recognize that these drivers are often correlated with, and drive, each other. Thus, a country that is ravaged by war and violence is more likely to have a weak legal system and be corrupt. Furthermore, all of these risk exposures are dynamic, and change over time, as governments change, violence from internal or external forces flares up.
As you assess these factors, you can see very quickly that country risk is a continuum, with some countries exposed less to it than others. It is for that reason that we should be cautious about discrete divides between countries, as is the case when we categorize countries into developed and emerging markets, with the implicit assumption that the former are safe and the latter are risky. To the extent that divide is not just descriptive, but also drives real world investment, both companies and investors may be misallocating their capital, and I will argue for finer delineations of risk…
… If your focus stays on economic risk, the question of whether democracies or authoritarian regimes are less risky for businesses to operate in depends in large part on whether these businesses are more unsettled by day-to-day continuous risk, which is often the case with democracies, where the rules can change when new governments gets elected, or by discontinuous risk, which can lie dormant for long periods, but when it does occur, it is larger and sometimes catastrophic, in an authoritarian government…
…In 2022, North America and Western Europe scored highest on the democracy index, and Middle East and Africa scored the lowest.
In my view, the question of whether businesses prefer the continuous change (or, in some cases, chaos) that characterizes democracies or the potential for discontinuous and sometimes jarring change in authoritarian regimes has driven the debate of whether a business should feel more comfortable investing in India, a sometimes chaotic democracy where the rules keep changing, or in China, where Beijing is better positioned to promise continuity. For three decades, China has won this battle, but in 2023, the battleground seems to be shifting in favor of India, but it is still too early to make a judgment on whether this is a long term change, or just a hiccup…
…When a country is exposed to violence, either from the outside or from within, it not only exposes its citizens to physical risk (of assault or death), but also makes it more difficult to run businesses within its borders. That risk can show up as costs (of buying protection or insurance) or as uninsurable risks that drive up the rates of return investors and businesses need to make, in order to operate…
…Iceland and Denmark top the list of most peaceful countries, but in a sign that geography is not destiny, Singapore makes an appearance on that list as well. On the lease peaceful list, it should come as no surprise that Russia and Ukraine are on the list, but Sub-Saharan Africa is disproportionately represented…
…Corruption is a social ill that manifests itself as a cost to every business that is exposed to it. As anyone who has ever tried to get anything done in a corrupt setting will attest, corruption adds layers of costs to routine operations, thus become an implicit tax that companies pay, where the payment instead of going to the public exchequer, finds its way into the pockets of intermediaries…
…Much of Western Europe, Australia & New Zealand and Canada/United States fall into the least corrupt category, but corruption remains a significant concern in much of the rest of the world. While it easy to attribute the corruption problem to politicians and governments, it is worth noting that once corruption becomes embedded in a system, it is difficult to remove, since the structure evolves to accommodate it…
…To operate a business successfully, you need a legal system that enforces contractual obligations and protects property rights, and does so in a timely manner. When a legal system allows contracts and legal agreements to be breached, and property rights to be violated, with no or extremely delayed consequences, the only businesses that survive will be the ones run by lawbreakers, and not surprisingly, violence and corruption become part of the package…
…By now, you can see the point about the correlation across the various dimensions of country risk, with the parts of the world (North America, Europe, Australia and Japan) that have the most democratic systems and the least corruption scoring highest on the legal protection scores. Conversely, the regions (Africa, large portions of Asia and Latin America) that are least democratic, with the most violence and corruption, have the most porous legal systems…
..Businesses and individuals that borrow money sometimes find themselves unable to meet their contractual obligations, and default, and so too can governments. The difference is that government or sovereign default has much greater spillover effects on all entities that operate within its borders, thus creating business risks…
…The most widely used measures of sovereign default risk come from a familiar source for default risk measures, the ratings agencies. S&P, Moody’s and Fitch, in addition to rating companies for default risk, also rate governments, and they rate them both on local currency debt, as well as foreign currency debt. The reason for the differentiation is simple, since countries should be less likely to default, when they borrow in their domestic currencies, than when they borrow in a foreign currency…
…One of the advantages of a market-based measure is that the market price reflects investor perceptions of risk at the moment. Sovereign Credit Default Swaps (CDS) offer a market-based measure of default risk, since investors buy these swaps as protection against default on government bonds. When the sovereign CDS market came into being a few decades ago, there were only a handful of countries that were traded, but the market has expanded, and there are traded credit default swaps on almost 80 countries in June 2023…
…The advantage of default spreads is that they provide an observable measure of risk that can be easily incorporated into discount rates or financial analysis. The disadvantage is that they are focused on just default risk, and do not explicitly factor in the other risks that we enumerated in the last section. Since these other risks are so highly correlated with each other, for most counties, it is true that default risk becomes an reasonable proxy for overall country risk, but there are some countries where this is not the case. Consider portions of the Middle East, and especially Saudi Arabia, where default risk is not significant, since the country borrows very little and has a huge cash cushion from its oil reserves. Investors in Saudi Arabia are still exposed to significant risks from political upheaval or unrest, and may prefer a more comprehensive measure of country risk…
…In addition to capturing risks that go beyond default, Political Risk Services also measures risk scores for frontier markets (like Syria, Sudan and North Korea), which have no sovereign ratings. The minuses are that the scores are not standardized… In addition, the fact that the country risk is measured with scores may lead some to believe that they are objective measures of country risk, when, in fact, they are subjective judgments reflecting what each service factors into the scores, and the weights on these factors. Just to illustrate the contradictions that can result, PRS gives Libya a country risk score that is higher (safer) than the scores it gives United States or France, putting them at odds with most other services that rank Libya among the riskiest countries in the world…
… For much of my valuation journey, the status quo in valuation has been to look at where a company is incorporated to determine its risk exposure (and the equity risk premium to use in assessing a hurdle rate). While I understand that where you are incorporated and traded can have an effect on your risk exposure, I think it is dwarfed by the risk exposure from where you operate. A company that is incorporated in Germany that gets all of its revenues in Turkey, is far more exposed to the country risk of Turkey than that of Germany.
3. Japan’s growing debt mountain: Crisis, what crisis? – Andrew Sharp
When the U.K. announced uncosted tax breaks last year, it triggered a run on the sterling, sent British government bond yields to their highest since the global financial crisis, and hastened the downfall of Prime Minister Liz Truss after just 44 days in office. This year, the U.K.’s ratio of debt to gross domestic product surpassed 100% for the first time since the early 1960s.
Japan could only dream of a figure so low.
The International Monetary Fund estimates that the world’s third-largest economy’s ratio is around 260% — by far the highest among developed economies, exceeding the 204% seen during World War II in 1944. The number is expected to continue creeping upward, according to projections by the Japan Center for Economic Research, a Nikkei-affiliated think tank.
Yet Tokyo remains relatively sanguine. In an optimistic scenario that sees a rise in Japan’s potential growth rate, the government projects it will balance its books by fiscal 2026.
The cost of borrowing, however, is rising. A decision by the Bank of Japan on Friday to allow yields on Japanese government bonds (JGBs) to rise above its previous cap of 0.5% to 1% has already triggered a spike in yields — they rose above 0.6% for the first time in nine years in Monday trading.
In the meantime, Japan keeps on spending. Prime Minister Fumio Kishida has pledged to boost defense expenditure to 2% of GDP by fiscal 2027 from around 1% now, and to double the child care budget to an annual 3.5 trillion yen ($25 billion). He is also planning to issue 20 trillion yen of Green Transformation (GX) bonds over the next decade.
While the GX bonds are to be repaid through a carbon tax and carbon pricing scheme, Kishida’s government has yet to settle on a plan to cover the defense and child-rearing outlays. Saddled with a super-aged society, the government projects Japan will have to spend nearly one quarter of GDP on social welfare such as nursing care and pensions in the fiscal year beginning April 2040.
So far, none of this has spooked global investors the way Truss’ tax plan did.
Various factors are dampening the fuse on Japan’s debt time bomb. Companies have large cash holdings and are not yet borrowing heavily. Japanese government bonds have a relatively long average maturity and are mostly held domestically. The country has a healthy current account surplus, and a rare period of inflation is also helping…
…Low growth as Japan’s population ages and falls is also a major risk. Without a significant productivity boost, a smaller working-age population would make it very difficult for Japan to maintain or boost growth, which would help to bring down the debt-to-GDP ratio.
“For Japan, the biggest social risk factor has been demographics,” de Guzman at Moody’s said.
4. Soft Landing Optimism Is Everywhere. That’s Happened Before – Jeanna Smialek
In late 1989, an economic commentary newsletter from the Federal Reserve Bank of Cleveland asked the question that was on everyone’s mind after a series of Federal Reserve rate increases: “How Soft a Landing?” Analysts were pretty sure growth was going to cool gently and without a painful downturn — the question was how gently.
In late 2000, a column in The New York Times was titled “Making a Soft Landing Even Softer.” And in late 2007, forecasters at the Federal Reserve Bank of Dallas concluded that the United States should manage to make it through the subprime mortgage crisis without a downturn.
Within weeks or months of all three declarations, the economy had plunged into recession. Unemployment shot up. Businesses closed. Growth contracted.
It is a point of historical caution that is relevant today, when soft-landing optimism is, again, surging…
…But it can be difficult to tell in real time whether the economy is smoothly decelerating or whether it is creeping toward the edge of a cliff — one reason that officials like Mr. Powell are being careful not to declare victory. On Wednesday, policymakers lifted rates to a range of 5.25 to 5.5 percent, the highest level in 22 years and up sharply from near zero as recently as early 2022. Those rate moves are trickling through the economy, making it more expensive to buy cars and houses on borrowed money and making it pricier for businesses to take out loans…
…That is not to say there isn’t good reason for hope, of course. Growth does look resilient, and there is some historical precedent for comfortable cool-downs.
In 1994 and 1995, the Fed managed to slow the economy gently without plunging it into a downturn in what is perhaps its most famous successful soft landing. Ironically, commentators quoted then in The Times weren’t convinced that policymakers were going to pull it off.
5. When did people stop being drunk all the time? – Lefineder
The English, said Sir John Fortescue (c. 1470), “drink no water, unless at certain times upon religious score, or by way of doing penance.”, looking at reconstructions of beer consumption from the middle ages to the pre-industrial era this was only a slight exaggeration. When estimating consumption from the amount of beer provided to soldiers, convicts, and workers or reconstructing consumption from tax revenues on beer we see that the average person consumed about a liter of beer a day, this is around four times as much as consumption in modern beer-drinking countries…
…Is this a historical overestimate? Probably not, in fact, there are several ways in which we might be underestimating historical consumption, most alcohol consumption in the past was from the local mono-drink but we should be still missing some amount of alcohol drunk by wine drinkers in the beer-drinking countries and vice versa and also the small amount consumed by spirits. In the medieval city of Ghent where there is data from 14th-century tax revenue on the consumption of both wine and beer2 per capita, annual consumption is:
- ~40-liters wine
- ~1300-liters beer (Such high figures are probably partly the results of the wealthy state of the city following the black death)…
…For English soldiers, it’s long been accepted to receive 8 pints of beer (4.5 L) as a daily ration9 an amount so great it probably was not wholly consumed, people did not have to use all their ration and they could also share it with their families. Nevertheless given that such quantities of alcohol were commonly supplied to historical armies the average soldier in the past didn’t just get angry for battle he got pissed. For sailors the beer supplied was of the strong kind (10%-15% alcohol) since this was the only kind that preserved itself well in the sea, hence drunk as a sailor. Such large consumption among workers and soldiers would mean that around a quarter to close to half of the calories in their diet were from booze…
…England transitioned to a low rate of beer consumption toward the end of the 18th century, looking at the more granular data on Malt beer consumption we see that this transition coincided with the timing of the onset of the British industrial revolution (1780-1800s).
Society is transformed in several ways, Whereas beer expenditure used to consume 12.5% of people’s salary in 1734 in the 1800s it consume only 1-3%. In the English poll tax of 1379-81 we can see that a total of 2.5% of the medieval workforce is comprised of brewers, in 1841 this is reduced to only 0.3 of the labor force…
…In the first of the following graphs, we see when people finish their workday, around 17:00. In the second graph, we see when people start drinking, the answer for the 18th-century cohorts is that drinking starts during the workday and already by 17:00 around 30% of people already drank liquor. In the 1830s this is no longer the case drinking on the job has seem to have been eliminated, people only start being recorded as drinking after 16:00. Society has been transformed by commercial forces.
Disclaimer: The Good Investors is the personal investing blog of two simple guys who are passionate about educating Singaporeans about stock market investing. By using this Site, you specifically agree that none of the information provided constitutes financial, investment, or other professional advice. It is only intended to provide education. Speak with a professional before making important decisions about your money, your professional life, or even your personal life. We currently have a vested interest in Alphabet (parent of Deepmind) and Meta Platforms. Holdings are subject to change at any time.