A growing number of scholars have expressed concerns regarding the societal, political, and economic implications of contemporary artificial intelligence (AI) trajectories. In The Anxious Generation, social psychologist Jonathan Haidt identifies a connection between social media engagement and rising rates of depression among teenagers. This concern arises as major tech companies prioritize user engagement to gather personal data for targeted advertisement. Similarly, researchers Acemoglu and Restrepo (2020) caution against a path leading to “the wrong kind of AI” and highlight the urgent need for ethical considerations and the protection of user privacy in AI development.
A recent paper explores the potential for redirecting AI development towards less data-intensive methodologies. Building on established literature regarding directed technological change, the authors argue that technological innovations tend to veer away from costly resources. Historical examples include automation technologies spurred by labor shortages and green technologies developed in response to environmental challenges.
Trends in AI Patenting
The paper introduces a framework to assess AI technologies based on their data intensity. While deep-learning methods require massive datasets to operate efficiently, knowledge-based systems and Bayesian methods rely more on structured rules and prior knowledge. Techniques like transfer learning and synthetic data generation are categorized as ‘data-saving’, contrasting with the ‘data-intensive’ nature of deep learning.
The authors demonstrate that since the introduction of the General Data Protection Regulation (GDPR), which increased the costs involved with managing personal data, there has been a discernible shift towards data-saving methodologies.
From 2000 to 2021, patents related to deep learning grew at an annual rate of 52%, while data-saving patents experienced only a 19% increase. However, following the GDPR’s implementation in 2018, the data-saving patent activity surged dramatically.
Privacy Regulations and Technological Innovation
The paper argues that GDPR has led companies worldwide to pivot toward more data-saving AI approaches. The employed analyses suggest that older and larger companies in particular were the primary drivers of this shift. Although this redirection may result in a decrease in overall patenting in the EU, it simultaneously emphasizes the market strength of established firms.
As AI has become increasingly intertwined with data-intensive methods, the authors underline concerns that privacy regulations could inadvertently benefit larger incumbents while stifling innovation. The upcoming EU AI Act may exacerbate this situation by adding regulatory burdens that small businesses find challenging to navigate.
Conclusion
The study emphasizes the potential for policy interventions to shape AI’s developmental trajectory significantly. As privacy regulations evolve, the balance between encouraging innovation and ensuring consumer protection remains a crucial area for future exploration.
