Takeaways - Notes From the AI Frontier MGI (Part II)

[Article 6 of 50. please refer to my first article for some background on this personal project and some standard disclosures. This is part two of MGI's Notes from the AI Frontier series. After discussing the insights from use cases last week, this week we look at the expected impact of AI on world economy.]

Here is the LINK to the full discussion paper, and here is the LINK to the summary article.

Introduction

As introduced last week, McKinsey Global Institute (MGI) published a series of discussion papers on Artificial Intelligence (AI) called “Notes From The AI Frontier”. After reviewing the use cases of AI in the paper I summarized last week, they examined AI’s impacts on world economy in this paper. The impacts were estimated part qualitatively (survey from government officials, corporation executives, thought leaders), and part quantitatively (proprietary simulation model). Their key findings are summarized in the section below:

1. AI could potentially deliver additional economic output of around $13 trillion by 2030, boosting global GDP by about 1.2% per year.
2. The impact will be slow at first, then accelerate after that.
3. The adoption of AI could widen gaps between countries, companies and workers.
4. The outcomes will depend on how countries and companies choose to embrace AI

Summary

AI could potentially deliver additional economic output of around $13 trillion by 2030, boosting global GDP by about 1.2% per year.

The above estimation is comparable to the other “general purpose technologies”. It was estimated that the steam engines increased productivity by 0.3% per year between 1985 and 1910; information and communications technology (ICT) and early digital technologies increased productivity by 0.6% per year during the 2000s.

The authors also broke down the economic impacts into seven dimensions (channels), and simulated the overall net impact by 2023 and 2030 respectively. The first three of these channels looked at how AI is enhancing productivity by augmenting human capital, replacing human capital and improving product and process innovation. The other four channels are external driven, they looked at the economic gains from improved global data flow, more wealth creation and reinvestment by countries and companies, as well as transition & implementation costs and other negative externalities. Out of all these channels, the author identified the gains from substitution, product innovation and the negative impact from AI driven competition and the resulting disruption as the major channels that affects global productivity. Another point that jumped out to me is the magnitude of impact by 2023 (+1% boost vs today) vs 2030 (+16% boost vs today). Which brings us to the next point.

The impact will be slow at first (the next five years), but accelerate after that.

Viewing this through historical lenses again, another general purpose technology, the aforementioned steam engine, was refined by James Watt in 1769, but it took until 1830 for steam to reach parity with water as a source of power in the British economy. The authors predicted that the aggregate net impact of AI won't really show in the next five to ten years. But they cautioned that companies should start to invest in AI now to reap the benefits down the road.

The adoption of AI could widen gaps between countries, companies and workers.

Countries:

In turn of readiness for AI, the authors grouped the countries into four groups:

1. Active global leaders (China and United States)
2. Economies with strong comparative strengths (noticeable countries include most of the Western and Northern European countries, Japan, Singapore and South Korea)
3. Economies with moderate foundations (noticeable countries include India, Malaysia and Turkey)
4. Economies that need to strengthen foundations (noticeable countries include Brazil, Indonesia and Zambia)

The authors argued that the gap between these different groups would increase over time.

The authors also ran the simulation model for eight economies in different groups to understand the impacts from different dimensions.

Companies:

Similar to countries, the authors grouped companies into three broader categories:

1. Front runners, 10% of population
2. Followers, 20-30% of population
3. laggards , 60-70% of population

The authors believed that front runners are set to capture most of the benefits.

Workers:

Previous MGI research estimated that by 2030, up to 375 million workers, or 14% of the global workforce, may need to change occupations. Some noteworthy points are listed below:

1. Jobs mix will shift toward high digital skills and not involving non repetitive tasks
2. Wage distribution will become even more uneven
3. The war for talents (with the appropriate skillsets) will intensify
4. In term of "robot taking away jobs" debate, the authors are surprisingly optimistic: they only projected a very small negative impact (-1%) for full-time-equivalent-employment (FTEs) by 2030

The outcomes will depends on how countries and companies choose to embrace AI

Finally, the authors proposed some questions for governments, companies and individuals to consider during this interesting time.

How countries can capitalize on AI

• How can countries step up investments that are beneficial in their own right but will also contribute to demand for work?
• ?How can policy makers evolve education systems and learning with a new emphasis on creativity, critical thinking, and adaptive and lifelong learning? How can countries increase investment in human capital, reversing the trend of low—and, in some countries, declining—public investment in worker training?
• How can policy makers improve the dynamism of labor markets, facilitate improved and faster matching of workers and jobs, enable a wider range of ways of working such as the gig economy, and solve such issues as portability of benefits, worker classification, and wage variability?
• How can countries embrace AI and automation safely, addressing issues including data security, privacy, malicious use, and potential issues of bias?
• ? How can countries rethink policy related to incomes to minimize disruption in case of a significant reduction in employment, greater pressure on wages, or both? Should they consider testing ideas such as conditional transfers, support for mobility, universal basic income, or some combination of the three? How can they best offer transition support and safety nets for workers affected? How can countries also forge a new social contract that garners support from all stakeholders including labor unions?

How AI will change the basis of competition among sectors and firms

• How are computing power and capacity, data, algorithms, and the availability of talent likely to evolve? How will developments in each of these affect individual firms? Which could benefit, and which might lose out?
• ? How can healthy competition be encouraged, maintaining an optimal balance in which front-runners are rewarded while minimizing the downside that could be imposed by winner-takes-all dynamics on later movers?
• ? How are industry structures likely to evolve, and how will sectors be redefined? For example, what role will be played by technology platforms that are most likely to be creator front-runners?
• ? What are the potential effects of widening gaps between front-runners and laggards, and how could these gaps play out in different sectors?
• How can companies redesign workflows to help workers adapt to working more closely with machines and fully absorb AI technologies across organizations? How can they create more collaborative, agile, and nonhierarchical organizations and cultures?

How individuals should be prepared for the AI-led transition

• How can individuals develop the skills that will be needed to power the AI economy and embrace a culture of lifelong learning?
• ? How can individuals leverage new ways of working, including participating in the gig economy and searching for jobs digitally?