The Progress Paradox: How Industrial Era Governance is Holding Back AI Era Growth

Introduction:

Over the past decade, though the global venture investment in Artificial intelligence (AI) grew over thirty fold – from $8.3 billion to $259 billion – the productivity growth in the United States stood at just 1.4% – much lesser compared to the 2.2% post-World War II average. Now, this poses a paradox – though it is presumed that the unprecedented technological capability unleashed by AI would result in a greater economic transformation, in reality it did not. The conventional explanation for this gap between AI’s enormous prowess and the lower realized impact focuses mainly on implementation lags: business reorganisation, worker reskilling and productivity measurement delays. While these could be partially true, the real answer could be in our governance systems, which belong to the industrial era, lagging behind in an AI era – thereby creating a bottleneck for progress.

Progress as Co-Evolution of Knowledge, Knowhow and Governance

Economic historians like Joel Mokyr proposed that growth is a result of the convergence, reframing and reinforcement between different layers. First is codified scientific knowledge – the knowledge layer, second, the human skills – the knowhow layer and third, the institutional frameworks – the governance layer. The Axial Age, a period between 800 to 200 BCE, is a great case in point to illustrate this historical interdependence and co-evolution of these three layers. The introduction of iron technology during this period – enabled progress in agricultural productivity, urbanisation, coinage and warfare – stimulating the knowledge layer. Simultaneously, these new ideas enabled the knowhow layer – where new agricultural practices and educational systems emerged. The governance layer, reflecting the spirit of the times, evolved in tandem – consolidation of kingdoms, emergence of empires, bureaucratic structures, standardised law codes and new forms of political legitimisation. These layers co-evolved – iron technology enabled bigger wars and higher agricultural productivity, which demanded better thinking and skills to manage, and which in turn required new governance structures to hold these together. Each layer reinforced the progress of others – resulting in an overall progress where all the layers are in synchrony.

The steam engine, invented during the Industrial Revolution further illustrates the same historical interdependence in its modern form. Initially deployed in the early 1700s, its widespread use did not simply limit the progress to steam power – but resulted in simultaneous evolution across all the layers: thermodynamics and engineering science (the knowledge layer), a crop of standardised, specialised and skilled factory workers (the knowhow layer), and new institutions – state bureaucracy, corporate law, capital markets (the governance layer) – ensuring the on-ground industrial era progress. None of these layers emerged in isolation – advances in the knowledge layer enabled better engine design, which demanded new skills from workers, which in turn required new state and labour laws to thrive. Each layer reinforced and enabled others – though not always at the same pace – but eventually adapted to allow overall productivity gains.

The Digital Era: Emergence of Institutional Friction as a Bottleneck

However, from late 1900 to early 2000 – the Digital Era appears to have broken this historical three layered synchrony. The introduction of automated software and computerised data capture accelerated the knowledge layer. Once technology sped up, new skills and competencies emerged and diffused through digital platforms. Historically, though the governance layer often lagged behind the others, this time around, it proved to be slower than the other times. It continued to rely on the industrial era processes: human driven bureaucratic rule-making, legislative cycles, economic protocols, and long drawn implementation cycles. Our current institutions, working at the industrial era timescale is facing enormous friction in processing the high speed production of knowledge and knowhow – creating a growth bottleneck. For example, while Moore’s Law halved computing costs every 18 months, the average time for the EU to adopt a legislation lengthened considerably: from 11 months in the 1999-2004 term to about 18 months in the 2014-2019 term – an increase of over 60% across two decades.

The AI Era: Compounding Institutional Friction

With the emergence of AI, humans are at a time and place where we can achieve immense unprecedented progress. AI has the ability to directly input digitised knowledge – scientific papers, models and datasets – as training data and create new models and insights at a breakneck speed. Similarly, AI also accelerates the knowhow layer – through workflow optimizations and skill simulations. For example, DeepMind’s AlphaFold discovered structures of 200 million proteins in a year, a task that would have otherwise taken centuries. However, in this AI Era, the asynchrony of the governance layer with the knowledge and knowhow layers just seem to have compounded. It is struggling to cope, absorb and regulate the complex highspeed proliferation of the first two layers. For example, the EU AI Act, proposed in April 2021, took more than 3 years to reach a political consensus – by which time the underlying AI technology had advanced several leaps forward – beyond what the legislators had originally contemplated.

Though AI is accelerating the first two layers, our current governance systems are still stuck in the industrial era – where regulations, rules and systems were written for environments that changed over decades – and hence would remain ineffective until the slow paced frameworks are reformed according to the times. Until then, the intellectual conditions artificially created by the AI, would far outpace the human created administrative and social conditions – creating a paradoxical lag in progress.

A Case for AI Assisted Institutional Design

In the AI Era, technology changes every few months, if not weeks. Regulations that take years to build become obsolete even before implementation. Hence, rather than following the industrial era based episodic institution building, we should now move towards a system where the innate strengths of AI are used alongside human validation systems – a system of algorithmic rule making where a stream of data is continuously input for live simulations and real time analysis, meanwhile preserving the fine balance of human oversight, approval and democratic systems. In fact, many of our systems like banking and finance and markets are already moving towards such agile and adaptive systems – to cater to the AI world. Such a system will help our regulators and legislators to cope with the increasing complexity emerging from the knowledge accumulation and knowhow automation processes – helping them identify conflicts, run simulations, reduce frictions and build early warning systems – at a pace more compatible with the AI driven economies.   

Conclusion:

While AI accelerates the layers of knowledge and know-how at a breakneck speed, the progress still lags behind, because the industrial era human deliberative governance layer slows down the overall interdependent co-evolutionary process between these layers. This asynchrony – not lack of technological ability is the real bottleneck of the AI era. Hence, the challenge of this era is in construction of intelligent institutions that reflects the spirit of the times – where the AI capabilities are used in building governance institutions in a human validated environment. And this might just unleash the true power of AI to move us towards a greater era of transformation.