Uncertainty is a constant
As a Quality Growth investor, you are often being whiplashed between Growth and Value. There are times when we are reprimanded for not being growthy enough. When optimism in high-level concepts abide, growth stocks run and our natural prudence and conservatism leaves us behind, we are confronted by the fantastic results of intrepid investors. At other times, when value stocks are performing well, we are chastised for the high multiples we pay for our investments, and for an alleged lack of valuation discipline that we displayed when times were good. Throughout it all we try to maintain an even keel, in the knowledge that there are market environments that are not best suited for what we do and where other investment styles will outperform us. In the last three years, we have witnessed the extremes of both these situations, together with a few others which defy categorisation. 2020 was a year which heavily favoured growth funds, while in 2022 value funds finally got their day in the sun. In both years we faced some criticism, albeit for diametrically opposed reasons. These are the normal challenges that investors who are true to their style will face. The key for us is to know what we are, communicate that clearly, and walk that narrow path.
In the markets, uncertainty, and the volatility that results from it are a constant. The salient question is what long-term strategy one applies to deal with this volatility. When looking ahead there is little to make us believe that this will change. There is much to be vigilant about as the market environment has not been getting easier. Liquidity continues to be a huge factor in determining the performance of assets, and there are things happening deep in the bowels of the treasury market that warrant close attention and could prove to be of some consequence. In this newsletter I will focus on artificial intelligence (AI), a subject that we have not yet written about but which will no doubt bring about its own interesting set of challenges.
The rise of AI
AI is a subject which has caught people’s attention in 2023. As far as people were aware of it at all, it was quite normal to dismiss the subject and be an AI sceptic, as there was much to be sceptical about. For many years artificial intelligence was a field marked by its numerous AI winters,1 where stagnation was the norm. Not too dissimilar to fusion energy, success was 20 years away, and always would be. This began to change with GPT-32 and really gathered momentum with the launch of ChatGPT when AI caught the awareness of the general public. The combination of Moore’s law, which over time has allowed a huge increase in computational power with the vast amounts of digital data that have been created in recent years has meant that previously discarded, and comparatively simple, computer models have vastly exceeded people’s expectations. The advent of cloud computing has also provided the technological backbone to make this widely available to the general public at a low cost. In a brief period of time AI has gone from being a field where relatively little appeared to happen, to one where it is near to impossible to keep up with the dizzying speed of developments. Superlatives abound with people comparing the impact that AI will have on society to anything from the invention of the steam engine, to the wheel and even fire.3 Increasingly, the standard view is that change will be revolutionary, not evolutionary.
The uncertainties of progress
Considering how little progress there was in artificial intelligence for so long, it may seem surprising that there is now such a sudden explosion in the rate of progress. It has caught everyone but the most optimistic by surprise, including those working in artificial intelligence itself. There is however much historical precedent for this, suggesting that the ability to forecast the rate of scientific progress is limited at best. Ernest Rutherford, the leading atomic physicist of his age, famously predicted in 1933 that anyone who looked for a source of power in the transformation of atoms was talking “moonshine”. Nuclear fission was discovered in 1938 and by 1942 Enrico Fermi managed to bring about the first self-sustaining nuclear reaction.4 In 1901, Wilbur Wright, one of the two Wright brothers, said to his brother Orville that human flight was at least 50 years away. By the end of 1903, the two brothers completed the first ever flight with the Wright Flyer. And Thomas Watson, Chairman of IBM is said to have estimated the global market for computers at maybe five computers in 1943. This uncertainty about the rate of progress continues to this day with experts holding widely different views on when Artificial General Intelligence5 will be achieved, if at all.
While the rate of progress is hard to predict, what is clear is that progress will continue and that its impact is likely to be meaningful. Even assuming no progress at all, which is highly unlikely, the impact of rolling out current AI models through the economy will be significant. However, while high level claims on the impacts of AI abound, such as the one I have just made, it is extremely difficult to predict what the impact will be in any concrete way. The technology is highly likely to be disruptive, but it is hard to say where that disruption will occur, who will be the ones doing the disruption and who will be the ones being disrupted. AI may well underpin the next wave of economic growth, but that growth could prove painful for those industries and professions at whose expense it is occurring, and it is near to impossible to predict how this will unfold. There is also no guarantee that the benefits of this growth will accrue equally across the economy, or that those who are currently leading the charge will be those who ultimately benefit.
Being roughly right and precisely wrong
Earlier in the year an internal memo written by a senior engineer at Alphabet was leaked. The memo put forward the argument that no-one really has a moat when it came to AI because the open-source community will quickly erode any advantage built up by individual companies. Not everyone agrees with this, but it raises interesting questions about how much we can say at this stage, in a concrete manner, about who the winners and losers of AI will be. Furthermore, most of the excitement surrounding AI currently revolves around Tech companies, be it those who are associated with the larger labs or those who are already rolling out products. But the impacts of AI will likely be felt further downstream. As the technology becomes more ubiquitous, the likelihood is that there will be many implications that we are currently not considering. Just as the rise of software and the internet brought about a proliferation of many new business models, the same is feasible for AI. Big picture claims and forecasts are currently being made, and these may well end up coming true, but big picture investing can be a dangerous road to embark on as detail and specifics matter.
In some ways the current environment shares similarities with the tech boom of the 90s. The advent of personal computing, the internet and telecommunications led to a flurry of excitement which ultimately provided the narrative underpinning the dot-com boom and bust. Despite the wave of enthusiasm we have seen this year, we are nowhere near the same level of excesses that we saw during the dot-com bubble, at least not yet, but there are a number of cautionary tales that can be drawn from that period. Notably, when one looks at the assumptions that were underpinning that time, they were largely correct. The internet has proven to be a ground-breaking technology and the infrastructure buildout that came with it was indeed gigantic. Combined with the ubiquitous proliferation of personal computers, it did allow for search engines to thrive and for e-commerce to put brick and mortar retailers under severe pressure. Many traditional business models went against the wall and new business models sprang up, providing a template for a fundamentally new way of doing business. Looked at with the benefit of hindsight, it is interesting to note that it was not the assumptions that were underpinning the dot-com boom where the mistakes were made. Yet vast amounts of money were lost because investors who got the big picture right, got the details and the timing very wrong.
Making broad predictions about outcomes is a very different proposition to making narrow forecasts about businesses and getting those right. The fact that a lot of infrastructure needed to be built out did not mean that the telecoms companies would be those benefitting, with the benefits ultimately accruing to an entirely different part of the value chain over time, such as companies like Apple and the end consumer. The fact that e-commerce was likely to explode did not help Pets.com who, despite identifying the trend correctly, failed in a spectacular manner.6 Google was not the first search engine, or even the tenth, but it is the one who prevailed. That gap between the conceptual story and the investment reality proved to be a painful lesson for many investors at the time. Similarly, companies like Amazon, Microsoft and Alphabet are spending billions of dollars on building out the computational infrastructure that will be needed for the widespread rollout of AI. Companies like Nvidia are the supplier of the chips that are best designed for neural networks to operate on. These companies are well placed to benefit based on what we know now, but this will likely be a fast-changing field. At this stage we do not really know who in the value chain will end up with the profits.
Despite having been around as an academic field since the 40s, AI is still in its infancy. The future of AI and the effects it will have on the economy, society and financial markets will be determined by a large number of factors and how they interplay with each other. Unfortunately, that means that they are going to be hard to predict. We can see that meaningful change is ahead of us, and people may be getting the big picture roughly right, but there is far too much uncertainty to make concrete investment decisions, at least from the point of view of the risk tolerance of a Quality Growth investor. Navigating this change as it unfolds will pose a challenge for investors and will require the ability to keep a level head. It is a trend which will play out over years, not quarters. Just like there are quarters where momentum favours Value over Growth, there will be quarters of AI optimism and quarters of AI pessimism. We have seen this already with the share price of Alphabet this year, with sentiment swinging wildly from believing that ChatGPT is a huge challenge to search to then believing that Google will be an AI winner, however that is currently defined. For us, the relevant question is not what we think of AI, but how will we deal with the unpredictability that it may bring about. It is, and will continue to be, a moving target. The key is to take the time to analyse each individual case in order to ascertain what can tangibly materialise, rather than placing conceptual bets. We will remain true to our Quality Growth investment philosophy and leave that to the pure Growth investors, who have the risk tolerance required for such bets.
For us, investing is not betting.
29th September 2023
1AI winter refers to a period of reduced funding, interest, and advancement in artificial intelligence caused in part by high expectations not being met. Historically, there have been three notable AI winters, during which the hype surrounding AI’s potential cooled.
2GPT-3 is a powerful language prediction model made by OpenAI. ChatGPT is based on GPT-3, but has extra training to make conversations better and easier to understand. It’s like a more user-friendly version of GPT-3 for chatting.
3In an interview with “60 minutes” Alphabet CEO said AI ‘is the most profound technology humanity is working on. More profound than fire, electricity, or anything we have done in the past’. (April 2023)
4Interestingly, when running this Newsletter through GPT-4, it tried to correct this statement, saying it was factually incorrect. As I disagree with GPT-4 on this, I have decided to risk it and leave it in. I will leave this to the physicists amongst you to judge.
5Artificial General Intelligence (or broad AI) is when machines can understand and learn any intellectual task that a human being can. This is different to Artificial Intelligence (or narrow AI), which is when machines can do simple or specific tasks like humans, for example recognising speech.
6Chewy.com, adopting a similar but better run business model, had net sales of $10bn in 2022.
Any forecasts, opinions, goals, strategies, outlooks and or estimates and expectations or other non-historical commentary contained herein or expressed in this document are based on current forecasts, opinions and or estimates and expectations only, and are considered “forward looking statements”. Forward-looking statements are subject to risks and uncertainties that may cause actual future results to be different from expectations. The views, forecasts, opinions and or estimates and expectations expressed in this document are a reflection of Seilern Investment Management Ltd’s best judgment as of the date of this communication’s publication, and are subject to change. No responsibility or liability shall be accepted for amending, correcting, or updating any information or forecasts, opinions and or estimates and expectations contained herein.
Please be aware that past performance should not be seen as an indication of future performance. Any financial instrument included in this website could be considered high risk and investors may not get back all of their original investment. The value of any investments and or financial instruments included in this website and the income derived from them may fluctuate and you may not receive back the amount originally invested. In addition stock market fluctuations and currency movements may also affect the value of investments.
Get the latest insights & events direct to your inbox
"*" indicates required fields