The AI Game

Regulating the future of work during AI Gold Rush?

This post is an excerpt adopted from a study “Transformation of work with AI”, developed as a part of GPAI project devoted to CAST framework. CAST supports responsible design and implementation of AI systems. The study benefits from creative insights and constructive critique provided by members of the CAST Project Team: Bogumi? Kamiński , Daniel Kaszyński , Robert Kroplewski and GPAI Future of Work experts: Dr. Saiph Savage and @Janine Berg. To read full study click on the link [...]

Also thanks to Jean-Luc Marcil Ferland for his editing support.

Introduction

Information technology has been transforming the nature of work for decades. Historically, computers evolved in two large application areas. One was industrial automation – controlling and monitoring complex machines in sectors such as transport, energy, or manufacturing. AI transforms this area through solutions such as autonomous operations and advanced analytics such as predictive simulation (digital twins).?

The second area, for which computers have always seemed to be particularly predisposed, is information management and knowledge work. Among factors specific for this kind of work is the importance of cognitive skills, such as: knowledge retention, comprehension, application, information analysis, synthesis, and creation [Bloom]. Hence a more general term – “cognitive work” – that may be applied to such activities. And information technology has been for decades providing and improving tools to make cognitive work more productive or – in case of computationally intensive tasks on large volumes of data – outright feasible. The process only gains momentum with the raise of AI becoming as a powerful “booster” for cognitive skills (see table 1).

Productivity paradox revisited

The growth of technology adoption in the last 4 decades, powered by exponential increase in computing performance and corresponding decrease of computing unit costs, seems to confirm the usefulness of information technology for many types of cognitive work. Yet, what puzzled economists looking into results brought by this growth was the “productivity paradox”, an observation attributed to Robert Solow, professor of economics at MIT who uttered in his book review published in 1987 by New York Times that “you can see computer age everywhere but in the productivity statistics”. Solow critical review inspired in depth research by scholars such as Erik Brynjolfsson, who has investigated the productivity paradox [Brynjolfsson] and remains an active researcher and leading authority in the area of technology economics, as well as practitioners such as Paul Strassmann , influential author of “Information Economics” [Strassman] who has developed methods to assess and manage economic value of IT investments with.?

This follow-up to work has led to some important insights. In 1998 study revisiting the productivity paradox, Brynjolfsson concluded that investments in organization (business processes and capabilities) have a considerable influence on the value of IT investments. He pointed out the fact that we are lacking proper tools to quantify and measure intangible benefits of IT investment such as increased competitiveness resulting from the fact that productivity gains are often shared with customers in the fight for market share and customer satisfaction. Strassmann, on the other hand, has demonstrated how productive application of digital innovations can be attributed to “Knowledge Capital” – a measure representing efficiency in running and managing operations. Strassmann, being an experienced CIO, who has managed IT organizations with budgets as big as $40 BLN has always advocated the view that IT solutions matter only to the extent, they make employees in operations more productive, reduce management overheads, and enable more competitive products. His views have been backed by analysis of the performance of organizations represented in the S&P index.

This paradox seems counter intuitive - certainly the capabilities that are at disposal of science, business, and society today are pure magic when compared to what technology was capable of 4 decades ago. And there are certainly examples of companies that are extremely successful in turning them into profitable growth. The thing is we tend to overlook the “whys”, mistaking tools for their innovative application.

The AI game

Monetization potential is one of main drivers behind AI growth. To achieve successful monetization investors and companies should revisit insights related to productivity paradox. They matter also at the verge of AI enabled transformation of work. There are 3 “whys”:

1. Computers don’t make money, people do [Strassman 2]. The relationship between IT spending and business results has been random for decades. Large IT budgets are no guarantee for becoming a more competitive or profitable. “How” the IT potential is used to increase topline or increase operations profitability by better management is what sets leaders apart from the pack. Brynjolfsson, Strassman and other researchers have demonstrated that technology innovations do not matter by themselves – they have to be integral part of business innovations – new organization, processes, value propositions and business models – to deliver results that matter. So transformation of work is not a by-product of AI implementation. It is (or should be) the most important rationale behind implementation.
2. We are at the beginning of the growth of a powerful technology investment bubble. Technological advances related to global infrastructures (such as rail, electric grid, creation of Internet, or Generative AI) generate investment cycles driven by inflated profit expectations and false assumptions about risk and opportunities. This cyclic nature of techno-evolution has been thoroughly analyzed by Carlota Perez in her "Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages". Perez has shown how over centuries emergence of new technologies and infrastructures, leads to frenzies of speculative investment and the formation of financial bubbles, followed by dramatic crashes. The productivity paradox has been just an early warning signal of a crash that culminated with the dotcom bubble burst resulting in investors losses estimated around $5 trillion between 2001 and 2002. As for AI, the global artificial intelligence market is expected to grow significantly by 2030. Statista predicts that the AI market will be worth nearly $2 trillion by 2030, up from nearly $100 billion in 2021. PwC's Global Artificial Intelligence Study estimates that AI could contribute up to $15.7 trillion to the global economy in 2030. AI investment vehicles are created also by public entities. The AI investment crash becomes a realistic threat.
3. Transformation of work by AI cannot be easily extrapolated. We cannot predict the scenario of work transformation from the existing experiences, just as we did not accurately predict many of the internet era transformations. Long-term market valuations show that the profits generated by winners have outweighed the loss of trillions of investment dollars. But the winners – companies such as Amazon, Apple, Google, AirBnB, UBER, Spotify etc – have disrupted established markets (media, retail, hospitality, transport), and redefined the business rules, models, and ecosystems in the way not foreseen by incumbent players.

Summarizing: the transformation of work with AI is a high-stake game from a business perspective and from the perspective of public interest. Winners emerging from this complex change might become extremely influential players, and assets contributing to the strength of national economies. At the same time this complex change combines the uncertainty that comes from business innovations (new products, business models and work processes) with uncertainties coming from technology innovations (new paradigms, architectures, tools, and practices).?

Conclusion - regulating the future?

Regulations may have a profound and long-lasting effect on society affected by technology. When the regulated markets are in the phase of an investment bubble, regulators are literally regulating the future, not ‘just’ addressing known issues and risks. The history of the internet provides examples of important regulations which have enabled the growth and proliferation of technology – net neutrality, privacy laws, or legal frameworks for online transactions. There are more than a few cases where regulation has been seriously lagging, struggling to cope with explosively rising issues such as the problem of online violence, or “computational propaganda” spreading dangerous misinformation in social media. Finally, overregulation, regulatory uncertainty or inconsistencies across jurisdictions can create barriers to growth. Some are justified by public interest as the barriers for multinational cloud providers operating in multiple regions, which have to invest in infrastructure supporting data residency and sovereignty regulations even if such infrastructure is redundant from the perspective of performance requirements. But the balance between regulatory friction and public value can be skewed by misunderstanding of technology, protectionism, or populism, becoming a modern version of the infamous “red flag traffic laws” .?

Harmonizing regulatory approaches and national policies for responsible AI needs to become a balancing act for the following aspects:?

1. Shaping an investment environment that enables the creation of national business assets – companies developing AI products and services which can be globally competitive, as well as provide incentives for companies to grow by applying AI innovations across their value chains. This includes enabling cross-domain and cross-border interoperability of data, algorithms and IoT infrastructure, especially related to critical autonomous systems (eg. transport, security, emergency response, critical grids).
2. Realization of the principles of Fair, Trustworthy, Responsible AI by providing incentives encouraging inclusion of these principles in investment strategies VC funds supported by public R&D and public investment funding institutions. Such constructive regulatory action can be more productive than prohibitive rules, monitoring, and penalization of AI misuse. ?
3. Manage investment (market) risks by inspiring and disseminating research of emerging best business and tech practices related to responsible AI. Ongoing education for investors and managers can help them shape the investment portfolios in a way that maximizes the chance for “walk of fame” – achieving market fit and successful scaling of products.
4. Finally, there is a need for AI practitioners to broadly adopt “responsible engineering” practices, learning from industries such as construction, aviation, or pharma. Such practices can be subject to hard regulation or soft law, including industry codes of engineering ethics, endorsed by national or international standardizing bodies, and industry associations.

References

[Bloom]: Adams NE. Bloom's taxonomy of cognitive learning objectives. J Med Libr Assoc. 2015 Jul;103(3):152-3. doi: 10.3163/1536-5050.103.3.010. PMID: 26213509; PMCID: PMC4511057.

[Brynjolfsson] Erik Brynjolfsson and Lorin M. Hitt, Beyond Computation:Information Technology, Organizational Transformation and Business Performance

[Strassman] Strassmann, Paul A. The squandered computer: Evaluating the business alignment of information technologies. Strassmann, Inc., 1997.

[Perez]: Perez, Carlota. Technological revolutions and financial capital. Edward Elgar Publishing, 2003.

[Moore]: Moore, Geoffrey A., and Regis McKenna. "Crossing the chasm." (1999).

[Statista] https://www.statista.com/chart/7428/45000-robots-form-part-of-amazon-workforce/