AI Unleashed and Other Platitudes

The equity markets overcame a slew of negative macro and fundamental developments and posted positive returns in the first quarter of 2023. In effect, equity investors seemed to be looking past the appalling trend in the ISM Manufacturing PMI most recent reading[1], lower IPO and merger activity, softening profit margins and earnings growth expectations[2] to conclude that the prospect of a near-term peak in interest rates was a catalyst for improving consumer and CEO sentiment, which would ultimately be conducive for long-duration assets, including equity markets. To be fair, investors are also likely looking past the shelter component measure of the US CPI (whose pitfalls we have highlighted before[3]) and assume that inflation is already much lower than the figure being reported. To be specific, the MSCI All Countries World Index[4], S&P 500 Index and the S&P TSX Composite advanced 7.02%, 7.36% and 4.55%, respectively, during the quarter.

Interestingly, excluding issuers who became takeover targets during the quarter, the best performers included a collection of ancillary cryptocurrency and artificial intelligence plays as well as some of the most significant large cap technology and consumer discretionary losers of yesteryear like Facebook (Meta), Tesla and Nvidia. The laggards, in contrast, were concentrated in the energy sector due to the decline in the price of crude oil and natural gas and, for reasons that we have already written about extensively earlier this month, the financial sector.

Fixed income markets rose in sympathy with equity markets as signs that higher rates were taking their toll on economy were multiplying, thereby sending bond yields lower across the curve. Movements in interest rate curves were particularly acute in the aftermath of the Silicon Valley Bank debacle as rates dropped by a record amount over a short period and implied volatility of interest rates, as proxied by the ICE Bank of America Merrill Lynch, surged to a record. While the US policy response, in particular the Federal Reserve’s Bank Term Funding Program (“BTFP”) was well perceived from a liquidity enhancing viewpoint, consensus is that the ongoing stress in the banking sector is leading to tighter lending conditions and the market’s response was to send credit spreads higher. Nevertheless, corporate bonds generally outperformed government bonds. To this point, the ICE Bank of America Global Government Bond Index and the ICE Bank of America Global Corporate Index returned 2,66% and 2,85% respectively. However, longer maturity bonds did much better than shorter maturity ones.

At the time of writing this, deposits in the US commercial banking system were still flowing out albeit at a slower pace than they did earlier in March[5]. As such, it looks like the regulators did the right thing. That said, I personally find it ironic that the venture capital industry, which was the principal recipient of the over-investment and over-lending largesse of Silicon Valley Bank and the likes, did not contribute a penny to bail-in the fragile First Republic Bank[6]. For the moment, it appears that the risk of a credit crunch has been contained as capital evaporation seems to have been largely limited to the venture capital industry, SPAC and other highly speculative sectors. It would be arguably more worrying if capital disappeared from the hands of the capital providers themselves. That would be more difficult to contain.

AI UNLEASHED

On November 30, 2022, artificial intelligence laboratory OpenAI[7] officially unveiled ChatGPT, a highly evolved prototype software designed to simulate conversations with human users. Built on the Microsoft Azure’s super-computing platform and released freely to the public, ChatGPT only required 5 days to reach 1 million users, the fastest ever application to reach such a milestone, faster than Instagram (2,5 months), Spotify (5 months) or Facebook (10 months)[8].

On January 23, 2023, Microsoft, which had already injected capital in OpenAI in two successive rounds in 2019 and 2021, made an unconfirmed additional commitment of 10 billion US$ to further its development. The announcement came less than a week after Microsoft launched an Azure service that provides access to a managed version of OpenAI’s ChatGPT[9].

The positive reviews garnered by ChatGPT’s release and its potential to transform Microsoft’s derelict Bing search engine reportedly caused tremors in Microsoft/OpenAI’s bay area rivals’ headquarters and it turbocharged a nascent AI race as Google and Baidu respectively launched Bard and Ernie during the quarter, their own response to the Microsoft/OpenAI partnership which had already come up with ChatGPT-4, an improved version of the original.

Let us take a step back to explain what ChatGPT is and why it matters.

ChatGPT is a conversational software that uses large sets of text data to learn to predict the most plausible next word, or sequence of words, based on the first few words of a sentence or a context. It is also capable of applying the knowledge acquired by solving a task to a related task. This type of large language models is called generative pre-trained transformers, hence the acronym (“GPT”). The data sets on which it was trained are vast, ranging from entire digital libraries, user manuals, research papers, newspaper articles and user-generated social media content.

The main advantage of ChatGPT over prior generations of conversational software is that it is extremely versatile. In effect, until recently, chatterbots were limited to provide answers to relatively narrow sets of questions. In contrast, ChatGPT is able to provide rapid, precise and coherent answers to complex questions in virtually every field. It is a revolutionary way to search the internet as instead of clicking through links for answers, ChatGPT does that exercise itself and returns the answer directly. ChatGPT is an indomitable Trivial Pursuit™ opponent. It is easy to understand why this newly available tool represents a serious threat to Alphabet’s ubiquitous search engine. But it goes well beyond that. ChatGPT remembers previous prompts given to it in a conversation and is thus able to fine-tune its responses based on prior questions. It can write long essays, summarize scientific articles, produce song lyrics, debug and create computer code and it is only a matter of time for it to be able to interact with other forms of media, like audio and video files.

There’s no denying that the speed and accuracy of generative Artificial Intelligence (“AI”) could boost labor productivity in many fields. In fact, in a recent report, Goldman Sachs[10] estimated that it could raise annual US labor productivity growth by nearly 1,5% per annum over a 10-year period following widespread adoption and boost Gross Domestic Product (“GDP”) by 7%. Goldman Sachs further estimates that 300 million jobs globally could be exposed to AI-automation. Interestingly, in contrast with the first wave of automation which displaced principally manufacturing blue-collar jobs, generative AI stands to impact white-collar jobs in multiple industries such as sales and marketing support, legal, education, accounting, design and engineering and many others. In some specific cases, generative AI is bringing the marginal value of human expertise down to nothing. From that standpoint, generative AI could be the most significant deflationary shock since China joined the World Trade Organization in 2001[11]. The spoils derived from productivity gains are typically not evenly distributed in society. In some industries, corporations might be able to retain the lion’s share of incremental profits. In others, consumers might disproportionally benefit through reduced cost of goods and services. Nevertheless, higher growth and lower costs have historically tended to be a positive combination for financial assets. For this reason, we are already incorporating the advent of generative-AI in our medium to long-term inflation forecasts and by ricochet, our medium to long-term asset return projections. To be clear, we still think inflation could surprise on the upside over the medium-term relative to consensus expectations but not as much as we thought a few months ago.

That being said, many people argue that generative-AI is not ready for prime time. They point to examples where ChatGPT failed to spot contradictions in a prompt, where it regurgitated misleading information, made defamatory claims or seemed to express certain political or cultural biases. Since ChatGPT is at its core a statistical model that makes inferences, it is normal that a small percentage of its responses will turn inaccurate. It will improve over time. Some critics are more fundamental. For instance, the iconic linguist and intellectual Noam Chomsky, in an interview with Open Culture[12] suggested that ChatGPT was nothing more than “high-tech plagiarism” and a “way of to avoid learning.” Along the same lines, songwriter Nick Cave, after having dissected lyrics written by ChatGPT in his own style simply said that it was “a grotesque mockery of what it is to be human.”[13] Others, alarmed by ChatGPT’s prowess go further, arguing that generative-AI has gotten close to the point where it poses a civilizational existential threat. In fact, over a thousand high-tech personalities have petitioned[14] AI research and development firms to pause their efforts to develop something superior to ChatGPT-4. I believe a voluntary pause is not a great idea because criminal organizations would never submit themselves to a voluntary pause. On the contrary, knowing that other generative AI developers were to submit themselves to a pause, the criminal ones would do everything to use the window to their advantage, the perfect recipe for AI development to go rogue. Anyways, I do not think that a development pause can be effectively implemented based on the historical adoption curve of some noteworthy revolutionary inventions. I note for instance that once Gutenberg rolled out his movable-type printing press, there was nothing that the masses of medieval monks who were tasked copying the Bible could do to prevent being disintermediated. Similarly, Karl Benz did not wait for three-way traffic light to make its appearance[15] or for Ralph Nader to push for the safety belt[16] to continue manufacturing cars safety belt[17] no more than Baltimore City authorities listened to gas lighting lobbyists after the incandescent electric lights were popularized in the 1880s, providing better quality light for one third of the cost[18].

Generative-AI has gone viral. Rightfully so as it offers a lot of promises. Notwithstanding that, it would be wise to remember what Sam Altman, ChatGPT’s creator, said about his offspring: “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness. It’s a mistake to be relying on it for anything important right now. It’s a preview of progress; we have lots of work to do on robustness and truthfulness.”[19]

With that in mind, it is unlikely that I will delegate to ChatGPT the task of writing future newsletters.

PROBLEMS FOR ZOMBIE AND VAMPIRE COMPANIES

Most of you have heard the term zombie bank or zombie company. It is used to describe companies that are generating cash and that are able to cover basic running and fixed costs, including the interest on its debt, but that would be structurally unable to repay the principal on their debt without bailouts, subsidies or dilutive equity financing rounds. The term zombie was borrowed from Haitian folklore and refers to a man who is under the effect of a filter or a charm that takes away his personality. Zombies are often portrayed as slowly decaying, mindless, soulless thralls without real purpose. The concept was first applied to many recipients of the support from Japanese banks after the bursting of the country real estate bubble in the late 1980s but was later applied to US financials in the aftermath of the Global Financial Crisis (“GFC”) of 2008 and to Chinese state-owned steel, aluminum and cement, and paper producers around 2015.

Corporate zombies have always been present in the public equity markets. From an economic standpoint, zombies have a negative influence because they inevitably consume more resources than they produce. If there are only a few, the negative economic effects remain manageable. But in a scenario where growth decelerates while debt servicing cost increase, zombies can quickly multiply and become a destructive force in the economy. At that point, the dilemma between maintaining zombies alive, by perpetually feeding them or letting them perish, is more easily resolved than when their number is a mere nuisance. To be clear, we do not think that a zombie apocalypse is imminent, but we believe that public policy towards zombie companies could become less complacent and their operating conditions might deteriorate rapidly. As such, existing and prospective zombie companies should rethink the long-term viability of their stupor.

Beyond zombies, there is another type of company has taken its name from a dreaded mythological creature. This type of company appeared more recently and is less common than the zombie company: the vampire company. The term was first introduced a few years ago by Professor Michael Wade[20] of the International Institute for Management Development (“IMD”) in Lausanne to describe companies that are able to offer their products and services more cheaply than the incumbents and offer unparalleled customer experience on a global scale. Like the vampire of the late romantic literary period, these companies are seductive and are relentless in the pursuit of their mission. This is the opposite to zombie companies. According to Professor Wade, vampire company examples include Uber (taxi and ride hailing industry), Amazon (bookstores), Netflix (movie rental) and Alphabet (internet search). Vampire companies operate by diverting revenues and profits away from industry leaders to themselves. To employ a vampiric analogy, they are taking over the blood pool.

The problem for vampires is that the blood pool ultimately runs out. For instance, the market share of Ford’s Model T went from 8% the year it was introduced in 1908 to 61% in 1921[21]. Although revenues continued to increase afterwards, growth decelerated because the pool was becoming empty and there were no more incumbents to displace. As it happens, the growth of vampire companies start to converge towards that of the broader economy. In the case of Alphabet and Facebook, the size of the blood pool is determined largely by the advertising budget of corporations who are paying to show up in internet searches. So, in order to survive, vampire companies need to tap into more blood pools, branch out into adjacent industries and categories. Some vampire companies have been incredibly successful at tapping into other blood pools. Amazon, who alone generates nearly 50% of all US e-commerce sales, is the vampire company overlord, in my opinion. Others have been less successful. In the past year, the top line growth of some of the better known vampire companies highlighted above disappointed and their stock price underperformed[22]. It is still too early to tell if this was just a blip or if the decline in the vampire companies’ stock price is a market’s call on the primary blood pool of these vampires running dry or worse, on their ability to find new blood pools, like the metaverse project for Facebook, the iCar for Apple or live-stream events for Netflix… What could get interesting is if two or more vampire companies were to set their eyes on the same blood pool. I wonder if a particularly fierce battle could cause a margin erosion to the point of turning both into zombies…

GöTTERDäMMERUNG[23]?

At the Davos Word Economic Forum in January, Mohammed al-Jadaan, Saudi Arabia’s Finance Minister, mentioned that the kingdom was open to settling trade for oil and natural gas in currencies other than the US dollar and the Euro. This statement drew a lot of attention, so I decided to investigate a little further to see if this indirectly signaled that the world reserve currency status that the US dollar has enjoyed since the Bretton Woods accord in 1944 was undermined. If so, it could lead to structural US dollar weakness for years to come and upend the Greenback’s role as a portfolio hedging tool for non-US domiciled investors.

Let us put a few things in perspective. First, consider that the US dollar’s share of the world’s central banks foreign currency reserves stood at roughly 56% in the third quarter of 2022[24]. It was approximately 70% at the turn of the century. The US dollar representation has been declining but the lost market share has been redistributed across a number of currencies, including non-traditional ones[25]. Second, consider that about half of the goods and service trade globally are invoiced in US dollars while the United States represents just 10% of the volume, both relatively constant figures over the last few years. Third, consider that according to the Bank of International Settlements, the US dollar is used in roughly 90% of foreign currency transactions (out of 200% as there are two currencies involved in a foreign exchange transaction). Also, a relatively constant figure in recent years.

At first sight, if an alternative currency is ascendant, it is not showing in the data yet. It is worth noting that China, who took over the United States as the largest trading partner of Saudi Arabia since 2011 (thanks to the US shale oil revolution which no longer made Saudi Arabia oil essential), has been pressuring the Gulf State to accept payments in Yuan for its oil and natural gas imports for a while. Incidentally, the two countries have gotten increasingly close recently. In fact, in the first quarter, Saudi Aramco acquired a 10% stake in the privately held Chinese refiner Rongsheng Petrochemical Co., Ltd. for about $3.6 billion and separately agreed to supply 490 000 barrels of crude oil per day to a Rongsheng subsidiary[26]. So perhaps it is just natural for the two countries to move away from the construct of a settlement currency convention between the United States and Saudi Arabia dating back to the 1970s. It is also possible, from a Saudi vantage point, that this is merely a way for the Kingdom to express that it does not regard its alliance with the United States as important as before, that it can prevail without the United States. Still, if the promiscuity between China and Saudi Arabia led to the settlement of crude oil exports to China in Yuan, what would it mean? Well, it turns out that the value of Saudi Arabia petroleum exports to China was 38,3 billion US$. By comparison, the value of the global trade was 32 trillion US$ in 2022. Therefore, the value of Saudi Arabia petroleum exports to China represented approximately 0,12% of global trade volume.

Based on the above, it does not seem that the US dollar term as the world’s reserve currency is coming to an end. It is likely just a slow-moving process by which central bankers aim to diversify foreign currency reserves and by which partners attempt to settle trade in their own currencies as opposed to an intermediary currency, something which has been made more possible with to the reduction in currency trading cost and other factors. Notwithstanding the above, it is possible that in a fractured geopolitical landscape, bi-lateral deals become frequent, and that the de-dollarization trend accelerates. But for the time being, the stability, depth and breadth of the US dollar and US dollar denominated assets coupled with the country’s military might and arbiter of conflicts history all lead me to believe that its demise is exaggerated. In fact, the US dollar index is only a few percentage points off last year’s peak. After all, Saudi Arabia might find some use for the Yuans it receives in exchange for oil, but I seriously doubt that the same approach can be extrapolated to the currency of trade partners like Indonesia, South Africa or Pakistan… If I am wrong and that demand for US dollars internationally falters precipitously, this could cause the borrowing rate of the United States to increase noticeably and for US imports to turn into a source of inflation for the first time in over 50 years.

Thank you for your support,

Dimitri Douaire, M. Sc., CFA
Co-Chief Investment Officer

[1] The University of Michigan indicator was at its lowest level since May 2020 at the end of February 2023.

[2] Source: FactSet

[3] https://patrimonica.com/rising-inflation-what-asset-owners-can-do-about-it/?lang=en

[4] Unless specified otherwise, index performance references total returns denominated in local currency terms.

[5] Source: Federal Reserve Bank of St-Louis, Weekly US Commercial Banks Deposits report.

[6] It was a consortium of banks led by JP Morgan that provided a 30 billion deposit backstop to First Republic Bank.

[7] Founded in 2015 in San Francisco by Sam Altman and a group of entrepreneurs including Elon Musk (Tesla, Space X, The Boring Company…) and Peter Thiel (PayPal, Palantir Technologies). The firm is also behind the DALL-E-2 and Whisper.AI applications.

[8] Source: Statista.

[9] Source: SiliconAngle.

[10] Goldman Sachs Economic Research, The Potentially Large Effects of Artificial Intelligence on Economic Growth, Hatzius and all. March 26, 2023

[11] Accession resulted in the reduction of tariffs and barriers to free trade.

[12] https://www.openculture.com/2023/02/noam-chomsky-on-chatgpt.html

[13] The Guardian, January 17, 2023

[14] https://futureoflife.org/open-letter/pause-giant-ai-experiments/

[15] Karl Benz is credited for the first functional four-wheeled automobile powered by an internal combustion engine in 1886. By contrast, the state of Connecticut was the first to install three-way traffic lights in 1930, 44 years later.

[16] Ralph Nader, Unsafe at Any Speed: The Designed-In Dangers of the American Automobile, Grossman Publishers, 1965.

[17] Ralph Nader, Unsafe at Any Speed: The Designed-In Dangers of the American Automobile, Grossman Publishers, 1965.

[18] The Illuminating Engineer, Public Lightning in Baltimore, 1909.

[19] Sam Altman’s twitter feed, December 10, 2022.

[20] https://www.imd.org/faculty/professors/michael-wade/

[21] Lacey, Robert. Ford: The Men and the Machine. Boston: Little, Brown, 1986

[22] According to Haver, the EV/EBITDA of companies involved in a digitization theme have derated between 28% (Healthtech) and 42% (Fintech) between November 2021 and October 2022.

[23] Webster: a collapse (as of a society or regime) marked by catastrophic violence and disorder.

[24] Source: International Monetary Fund estimates

[25] Other than the Euro, the Japanese Yen, and the British Pound.

[26] Source: Reuters.