For decades, particularly in advanced economies, productivity growth has been in a prolonged downturn.
This has been a major challenge for policymakers, a source of anguish for businesses, and a drag on global GDP.
Despite plenty of targeted investments in physical and human capital, the establishment of supporting collaborative ecosystems to enable innovation, and resource allocation to the most productive sectors, the productivity problem has remained a significant challenge.
Source: The Conference Board Total Economy Database
However, this key obstacle is now being challenged, thanks to the promise of Artificial Intelligence (AI).
THE FUNDAMENTALS
Broadly, AI is an umbrella term for a set of technologies which can enable powerful computer systems to sense their environment, learn, and respond according to stimuli.
It attempts to impart human-like intelligence to advanced computer systems.
With such technology, the computer’s superior ability to identify patterns, explore masses of data at an unprecedented rate, and identify and evolve its own understanding of scenarios, is a truly remarkable advantage.
AI carries the incredible potential to alter global productivity trajectories by releasing gains which were impossible only a decade earlier while unlocking completely invisible revenue streams.
Although many techniques have been well-known for decades, the generation of incredible volumes of data, a relatively new but exponentially-growing phenomenon, has enabled a true revolution.
In an unexpected twist, the outbreak of the global health pandemic served to greatly accelerate the process of innovation, as firms were desperate to secure an edge over their competition in a fast-changing environment, while the bulk of the workforce turned to Work-from-Home arrangements.
TL;DR
- Artificial Intelligence (AI) is a set of technologies that give computers human-like intelligence to learn and identify patterns.
- AI promises to boost productivity growth which has been lagging in major economies for many years.
- The algorithms have seen huge improvements in recent years due to the explosion of data availability.
- In 2021, corporate investments in AI technology reached historic highs, driven by the global health crisis.
- The global AI market is valued at well over a $100 bn and is forecast to see tremendous growth this decade.
- The incredible adoption of ChatGPT and similar programs has revitalized the sector after a slowdown in 2022.
- Investments, adoption, and productivity growth will vary across industry and between countries.
- Researchers have estimated that 85% of new jobs since 1940 in the US were created due to evolving technology.
- Central banks are likely to become major adopters of AI due to the need for better data-driven decision-making.
- Monetary policymakers are yet to fully understand the role that AI may play on major economic variables such as inflation, wage growth and interest rates.
- AI is democratizing investor education, key insights and portfolio optimization among retail investors.
THE SIZE OF THE MARKET impacted by ai
We create an incredible amount of data each day, and in most cases, tapping into that data allows AI models to be better trained and more adaptable.
Estimates suggest that in 2022, 333 billion emails were sent out each day.
Users tweeted 650 million times in any 24-hour period, sharing a wealth of information, perspectives, and unique insights.
Each of these interactions offers invaluable data for AI robots to sharpen their processes and outputs.
By the year 2025, it is forecasted that over 200 zettabytes of data will be stored in the cloud.
To put that into perspective, one zettabyte is a trillion gigabytes.
The sudden rush of unique data is yielding new opportunities in the AI space, not only for ambitious entrepreneurs and companies but for the economic system, as a whole.
A Forbes report noted that in 2022, the global AI market was valued at a staggering $137 billion, and is projected to grow at a CAGR of over 37% over the next 7 years.
A PwC study expects AI to add an incredible $15.7 trillion in terms of global GDP by 2030, with far-reaching impacts in China and North America, equating to a boost of 26% and 14.5% of GDP, respectively.
Source: PwC
A McKinsey and Company research study estimated that through adoption in areas such as manufacturing, healthcare and finance, global GDP could be boosted by an additional 1.4% to 2.6% by the year 2030.
The following diagram shows the variety of AI technologies that have already become mainstream across sectors, as of 2022.
Source: McKinsey and Company; Stanford University
For instance, Robotic Process Automation is the most likely AI technology to have been embedded across all industries at 39%.
RPA is deeply embedded in Business, Legal and Professional Services (46%), Financial Services (47%) and High Tech/Telecom (48%). Across all industries, Computer Vision, Natural Language Text Understanding (NLU) and Virtual Agents have also seen widespread adoption at 34%, 33% and 33%, respectively.
Yet, there are significant efforts to develop novel data flows that can also tap into both structured and unstructured data from unique sources.
A quick word of caution to our community of readers – AI is continuously changing and how its impacts will be shaped in the future depends on several ever-evolving factors.
So, what does all this mean for the wider economy? Let’s take a look.
CAPITAL INVESTMENTS into AI Technologies
Source: Our World in Data
In 2021, global corporate investments in AI peaked at record highs of $276 bn, an incredible 79.7% increase over 2020.
Much of this value was unlocked via private investment, which accounted for 45.4% of the total outlay, and M&A which made up 43.4%.
Of course, this huge injection was partly due to the sudden shift of WFH and the heightened demand for digital goods and services.
Unsurprisingly, there was a fall in investment in 2022, declining to $175 bn, marking a decline of 36.5% over the previous year.
To put this into perspective, the total investment in 2021 was greater than the combined reported value in the six years from 2013 to 2018, while 2022 was larger than the five years from 2013 to 2017.
Parallelly, the nature of technological adoption is that it largely follows an S-curve, which means that adoption starts extremely slow, and picks up pace until it reaches a critical mass, after which it explodes higher.
It is only once adoption becomes near complete, that this begins slowing significantly.
A study from Harvard Business Review shows this dynamic for several technologies in the past and can be made out for some technologies such as the microwave, computer, and the internet.
Source: Harvard Business Review; The New York Times
If we consider AI to be an umbrella term for component technologies, the incredible speed of adoption for ChatGPT for instance, may be signalling that we are now approaching the accelerated portion of the curve.
This, in turn, is likely to lead to a reinvigoration in AI investments in the coming years.
A data spike in companies mentioning AI on earning calls further supports this.
Source: Stanford University
impact of ai #1 - PRODUCTIVITY
As mentioned earlier, increased productivity is one of the key outcomes that embracing AI is likely to deliver.
These gains are expected to come through two separate channels.
The primary channel would enable more efficient work in tasks that can be automated.
Industry estimates suggest that in most cases of roles that are potentially exposed to AI, a quarter to half of the tasks may be automated.
A secondary channel would improve output in the aggregate by creating demand for labour in new emerging industries that may have been displaced by AI.
For instance, the rollout of the PC enabled a variety of new professions such as web designers, software engineers, and digital marketing specialists.
A variety of reputed academic studies have estimated the impact of AI adoption on annual worker productivity growth to lie between 2%-3%, while some researchers estimate this to be closer to 7%.
Meta-studies indicated a median productivity increase of approximately 2.6% per year.
This uncertainty is due to still-emerging business models, varying timelines for adoption, the varying complexity of tasks that may need automation, and the efficiency of upskilling initiatives.
Analysts at PwC estimate that labour productivity improvements are likely to account for a massive 55% of additional GDP generated by the year 2030.
This is supported by an NBER study published in April 2023, which found a 14% improvement in ‘issues resolved per hour’ by the introduction of AI-based conversational tools in customer support situations.
As per Goldman Sachs, over a 10-year period, global annual productivity is forecast to experience a 1.4% increase in annual productivity.
Source: Goldman Sachs
Due to easy access to digital infrastructure in advanced economies, these countries are likely to see higher productivity gains due to job automation, especially in the near term.
A word of caution to our readers is that historic data on ‘milestone technologies’ have delivered enormous productivity gains, but studies show that the extent of these gains can and the time taken to be realized may vary.
Source: US Bureau of Labor Statistics, Census Bureau, Our World in Data, Woolf (1987), Haver Analytics, Goldman Sachs Global Investment Research
impact of AI #2 - Changing work
Although labour productivity is necessary for the long-term revival of economic growth, there are several concerns relating to the impacts AI could have on the job market.
In particular, will AI adoption lead to huge job losses?
Fortunately, historical analysis of earlier far-reaching technological revolutions does not support this conclusion.
However, studies indicate that although aggregate jobs may not decline, the nature of work is likely to change dramatically.
This was seen at least twice during the twentieth century, in the case of the modernization of oil, rail and automobile industries; and later, during the early computerization of the 1980s.
As discussed in the section on productivity, in most cases, worker displacement was offset by the creation of new jobs in related or secondary areas.
A 2022 study led by Professor David Autor of MIT, used an interesting approach to study the emergence of new roles and employment amid technology-driven growth.
As per their research, over 60% of employees today work in professions that did not exist in 1940, the majority of which are thought to be due to new technological advances.
Source: Autor, D. et al. (2022)
The blue bars depict employment levels in 1940 according to occupation.
The green bars are the jobs that are present in 2018 and were available in 1940 as well.
The red bars, which are especially pronounced in Professionals, Managers, Technicians, Clerical and Administrative Work, Construction and Health Services, are jobs that did not exist in 1940.
Goldman Sachs research notes that this implies that,
“85% of employment growth over the last 80 years is explained by the technology-driven creation of new positions.”
Source: Visual Capitalist
It is important to note that the above estimates relate to tasks within professions, implying that a quarter of all tasks across all industries could be automated.
46% of administrative support tasks may be vulnerable to automation, potentially boosting productivity in this area.
Similarly, for the euro countries, research indicates an aggregate estimate of 24%, with clerical staff tasks being as high as 45%, but this was as low as 4% for trade workers and craftsmen.
In the long run, industries that draw on heavy amounts of data are likely to see more disruption and benefit from higher productivity gains, such as in the case of retail, healthcare, or finance.
Globally, estimates suggest that 18% of tasks could be open to automation.
Source: Goldman Sachs
Overall, emerging markets lag due to the difference in economic structure as against countries such as Singapore, the UK, and the US.
Agricultural positions, for instance, are not as open to automation as those of advanced economies.
However, the attitudes of adopters including business owners and employees can also vary significantly.
A survey by Llyod’s Register Foundation highlights these differences, where attitudes towards AI in China were positive at 4.5, while in the US, they were somewhat negative, at 0.9.
The value of 4.5 indicates that 4.5 people are optimistic about AI for every person who expected it to cause more harm.
impact of AI #3 - SKILL BIASED TECHNOLOGICAL CHANGE
In the case of most technological shocks, adapting to this new reality may demand that workers upskill to meet the requirements of new and changing roles, which is referred to as ‘Skill biased technological change’ or SBTC.
Without upskilling, workers may face a fall in real wages.
By and large, the literature finds that workers who adapt best to their new competencies tend to earn a premium, but this depends heavily on the industry and country of residence.
Echoing the S-curve, Darrell West, Senior Fellow at the Center for Technology Innovation at Brookings Institution, noted AI adoption could be much faster than widely anticipated.
He suspects that prompt-driven, generative AI tools could potentially transform job roles in most industries within a 2-to-3-year period.
For instance, ChatGPT was reported to have a near-vertical adoption rate, with considerable efforts being made to integrate this technology into workflows.
Source: John Nosta
impact of AI #4 - CENTRAL BANKS
Monetary policymakers are facing unique challenges today.
After a prolonged period of low interest rates following the GFC, the unforeseen global health crisis triggered four-decade highs in inflation.
Central banks scrambled to respond, and although inflation has come down significantly, it remains above target, and the US labour market has been more robust than anticipated.
To address such highly complex and evolving challenges, monetary authorities require access to sophisticated tools and the ability to finely analyse ever-growing sources of information, often in real time.
4a. AI-LED INFLATION, REAL INTEREST RATES AND MACROECONOMIC CHALLENGES
A crucial consideration for central banks will be to monitor and understand the role that AI adoption may play in inflation trajectories and real interest rates.
The effects will vary across countries, and depend on labour market shifts as well as public attitudes towards adoption.
At the outset, we may expect higher productivity of AI to lead to more investment avenues and higher demand for capital, thus, raising real interest rates (Rate of Interest minus Rate of Inflation).
At the same time, higher productivity may bring down inflation through lower unit costs.
Having said that, several ambiguous effects may potentially emerge.
The concentration of capital generated, impacts on savings rates, labour participation rates, changes in retirement age, the pace of adjustment of wages, and resultant impacts on total demand, could all contribute to determining how real interest rates and inflation change.
A crucial concern for data-driven decision-making would be the emergence of slow adjustment by the labour force amid SBTC-type dynamics which could lead to lower bargaining power and thus, lower real wages and increased vulnerability to recessionary risks.
In terms of market concentration, which the Brookings Institution noted was associated with greater digitalization, a May 2023 Bloomberg article found that the S&P 500’s AI-fuelled rally has been extremely narrow.
Source: Bloomberg
In fact, the Big Seven’s median gain since January was almost five-fold greater than the entire market.
Moving ahead, monetary authorities will need to track any knock-on effects on consumption and expenditure and modify their view of the health of an economy while approaching market interventions.
AI offers a rare opportunity to kickstart productivity growth after prolonged sluggishness, but will equally require authorities to monitor imbalances while not inadvertently stifling innovation.
Industry forecasts suggest that business investment in equipment and intellectual property is likely to remain subdued in the medium term due to the prevailing higher rates policy.
In addition, markets continue to face an inverted yield curve, as well as challenges in the commercial real estate and residential sectors.
These factors could in turn lower demand and reduce purchasing strength.
4b. ADOPTION AND CASE STUDIES
In a bid to improve decision-making, the Bank of International Settlements (BIS) found that the use of Machine Learning (a subset of AI) and big data ballooned from 30% of central banks in 2015 to 80% in 2020.
This spurt was also reflected in the growing use of the term “big data” in speeches by monetary policy officials and continued to take on a disproportionately positive tone.
Source: BIS
Source: IFC (in % terms)
Source: BIS
The Bank of England jointly with the ECB developed clustering techniques that use over 6,000 time series data sets, spread across 30+ countries to quickly identify potential economic shocks and any worrying anomalies.
The Bank of Indonesia developed a system based on a host of techniques such as random forest, SVM and XGBoost to analyse 2,000+ variables, and thus, identify the behaviour of foreign investors vis-a-vis local debt markets, and rapidly adjust policy.
Another interesting case emerged from the Central Bank of Malaysia, which developed an AI tool to improve communications and ensure consistent messaging, a crucial component of expectations management.
However, challenges remain.
In the OMFIF’s 2023 Global Public Investor Survey, ‘recruitment and training’, ‘access to effective analytical tools’, and ‘access to relevant data or information’ were the top 3 operating challenges faced by central bank reserve managers, pointed out by 58%, 49% and 33% of respondents, respectively.
impact of ai#5 - POTENTIAL CHANGES TO INVESTING
Learning from large datasets enhances data analysis, improves risk assessment, optimizes portfolios, and perhaps most importantly, mitigates investor bias to provide better decision-making.
Institutional investors such as BlackRock and Bridgewater have invested heavily in AI and big data capacities to streamline investment processes and better gauge market reactions.
In addition, generative AI platforms and neural network architectures are democratizing detailed insights that were previously unavailable to individual retail investors.
Such advances in AI are making it possible for the common person to adopt a more scientific approach to their investing.
Again, these technologies may have truly remarkable implications on various facets of investing and the macroeconomy given the prolonged period of low to falling productivity.
FINAL WORDS
AI technologies are rapidly evolving, boost global productivity and are forecast to contribute meaningfully to GDP.
In 2022, we collectively sent over 650 million tweets a day.
In 2022, the global AI market is valued at $137 bn by PwC analysts.
So far, RPA is the most implemented AI technology across all types of businesses in the US.
- Over a 10-year period, AI technologies are expected to improve global productivity by 1.4% per year.
- Office and administrative support, legal work and the architecture and engineering professions are most amenable to wide-scale automating of tasks.
- Skill-biased technological change (SBTC) is likely to play a major part in future labour markets and job matching.
- Central banks are rapidly adopting AI tools to improve decision-making.
