SG Kleinwort Hambros CIO Blog November 2023
“I, for one, welcome our new computer overlords.” Such were the words of Ken Jennings, legendary contestant on the US gameshow “Jeopardy!”, after being de-throned by IBM’s supercomputer “Watson” in 2011. Now, little more than a decade later, a new form of artificial intelligence (AI) has taken the world by storm; one that is not only more powerful but also more easily accessible. This new paradigm of so-called generative AI such as ChatGPT is still in its infancy.
New models come remarkably close to what one could consider sentient but are still far away from an elusive “artificial general intelligence”. They largely parrot what’s been said and written before. They struggle to generate truly new, innovative ideas. They get things wrong – but so do humans. The current capabilities of popular AI are more akin to those of a rather capable personal assistant. In the context of white-collar work environments, they can research, draft, summarise and edit, help brainstorm and conceptualise.
Any ideas and creative input must still be generated by humans, but the heavy cognitive lifting can be handed over to the machine. We are only beginning to see the vast range of applications for this new technology across virtually all parts of the economy: from predicting the shape of proteins in medical research to creating bespoke personal DJs on Spotify. Nonetheless, the technology is almost certain to inject a dose of what economies, in particular those that are developed, are in dire need of: productivity growth. Labour productivity is loosely defined as the amount of output a given worker can produce per hour worked. Artificial intelligence can support humans in multiple ways: in many instances, it will support workers to produce more in a given amount of time or automate parts of the workflow altogether. In other cases, workers will be able to produce not more, but higher-quality products. Supported by AI, leaders will be able to make better decisions to improve corporate strategy and reduce waste.
Bank of America estimates that labour productivity will rise by as much as 40% until 20351, translating to about 3% per year. This would be a significant acceleration from the sluggish growth many developed economies experienced since the turn of the century. Labour productivity in the UK grew at an average rate of 0.7% over the past 20 years2; the US, a global vanguard of technological innovation, managed 1.5%. Indeed, the impact of artificial intelligence may well exceed that of the last major technological revolution: productivity grew 2.2% to 3.0% in the years between 1999 and 2005, just around the time many enterprises began to utilise the benefits of the world wide web. Others are not quite as optimistic: analysts at Goldman Sachs reckon that3, even if artificial intelligence will provide a short-term burst of productivity growth, this will be more than offset by a structural slowdown in economic growth over the next two decades.
With great technological change of course comes great concern. Will humans finally be replaced by machines? Will a new wave of automation give rise to wide-spread unemployment? Some roles will naturally disappear as technology advances and new jobs will be created to do work that has hitherto not existed. Generally, however, history has shown that fears of structurally higher unemployment are largely misplaced. These concerns rest on what economists call the “Lump of Labour Fallacy”: the mistaken belief that a set amount of work in a given economy is distributed among a varying number of workers. This framework suggests an economic zero-sum game, in which any increase to the available labour force (or their productivity) would lead to a rise in unemployment. In reality however, it is much more likely that enterprises are able to utilise additional resources to supercharge growth.
Consider the tech industry: if a programmer – supported by AI – can produce 40% more code in a decade from now, are we more likely to see 30% fewer programmers or 40% more output? In a world driven by an insatiable demand for innovative software, we are much more likely to enjoy a faster supply of higher-quality applications with improved customer service to match. Of course, this progress is likely to affect separate parts of the economy to different degrees: benefits will be more concentrated in developed economies with larger services sectors and higher investment in research and development. Even within countries, sectors with existing IT infrastructure and talent will be faster and more efficient in adopting new processes. Nonetheless, as demographics continue to deteriorate and the labour force becomes structurally tighter, any efficiency gains will be crucial to support ongoing economic expansion.
Despite the latest hype around the wonders of artificial intelligence and the companies that enable it, we expect the actual transition to an AI-augmented economy to be somewhat slower and more gradual than widely assumed. Firstly, the supply of resources required to build and train capable models is already under serious strain. The largest makers of semiconductors are struggling to meet demand for specialised chips. Once assembled, powerful data centres consume vast amounts of energy, prices of which have skyrocketed amidst rising geopolitical tensions. Compounding this, is the movement to net zero and the complexity of achieving a sustainable energy mix. Ultimately, these factors are beneficial for enterprises on the supply-side of the transition, whose order books are jam-packed for years to come, however, will limit the pace of large-scale adoption. Secondly, the forthcoming technological disruption will be accompanied not only by economic uncertainty but risks to national security, politics, and society at large. Governments worldwide are eagerly consulting with industry leaders on crafting legislation, with President Biden just recently issuing a far-reaching executive order4 to limit the potential harm from artificial intelligence.
The goal is to both mitigate significant risks and prevent an undue concentration of power, all while preserving the pace of innovation. It is a complex endeavour, and it will take some time yet. Finally, much depends on the willingness and capability of enterprises to transition to new processes and technology. With cost pressures set to persist, many executives might be averse to initiating expensive IT projects. Leaders and workers may be reluctant to change, embracing established yet suboptimal processes over new ones with uncertain outcomes. We-have-always-done-it-this-way culture may prove to be one of the biggest obstacles, and adoption laggards are likely to find themselves outpaced by those enterprises embracing change.
A gradual transition will ease the potential pain for workers as some legacy jobs make way for roles in a new paradigm. On the other hand, it is likely to pace the realisation of economic benefit such as productivity growth over a longer time period. Either way, a new paradigm awaits, and investment opportunities abound.
The return of the US stock market in 2023 has been driven to a significant degree by the “Magnificent Seven”5, a collection of tech stocks comprising of Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta, and Tesla, that rallied hard on the back of the AI hype unleashed by the emergence of ChatGPT last year. At the first glance, parallels to the 2001 dot-com bubble are evident, however, they are largely unfounded. Two decades ago, investors were emboldened by a steadily loosening US monetary policy and as such, investing into internet companies with big ideas but little profits became the norm. Today, we are seeing much of the opposite: enthusiasm for high-quality, profitable enterprises - albeit at high valuations – in an environment of rapidly tightening monetary policy.
The ship for AI investing has not yet sailed. We are likely witnessing the beginning of a new paradigm that will shape the economy for decades to come. It is impossible to forecast individual winners in advance, and first-movers do not always endure: even behemoths such as Google or Amazon that are now near-synonymous with their respective industries were, in fact, late entrants to their markets.The thematic can be played across the value-chain, starting from energy suppliers and chipmakers, to cloud infrastructure, software developers, and finally users. Especially the latter category provides a deep universe of investment opportunities, as AI has the potential to disrupt business models across all sectors of the economy, streamlining supply chains of major retailers, supporting minerals exploration for miners, or ensuring perfect cheese distribution for frozen pizza manufacturers.
Identifying winners is easier said than done: the use of the words “artificial intelligence” surged to record heights in 2023 earnings calls but few enterprises can point to meaningful disruption just yet. We prefer exposure to markets with an established history of fostering – and capitalising on - innovation, such as the US, and are looking for quality businesses with sustainable leverage and profits to build on. Indeed, a company’s proven track record for change provides a solid investment case, especially if it is supported by strong IT capex and leadership. Finally, we recognise that the impact of artificial intelligence will not materialise overnight. Hence, we prefer longer investment horizons over opportunistic trades of popular names. After all, this journey is only just beginning.
1 AI Trends Report: Industry Impact of Artificial Intelligence (bofa.com) 08/03/20232 OECD, Kleinwort Hambros 31/10/20233 Explaining the Productivity Lull (Hill/Abecasis) (gs.com) 29/10/20234 Biden Issues Executive Order to Create A.I. Safeguards - The New York Times (nytimes.com) 30/10/20235 Performance sourced from Bloomberg, 31/12/2022-31/10/2023: Apple +32.0%, Microsoft +41.0%, Alphabet +40.6%, Amazon +58.4%, Nvidia +179.1%, Meta +150.3%, Tesla +63.0%. Past performance is not an indicator of future performance.
