In recent discourse surrounding artificial intelligence (AI), increasing concerns have emerged about the social, political, and economic ramifications tied to its ongoing evolution. Scholars like Daron Acemoglu and others have expressed apprehension regarding what they term “the wrong kind of AI.” This sentiment is echoed by social psychologist Jonathan Haidt, who, in his work "The Anxious Generation," connects the rise of social media usage to a rise in teenage depression, contending that larger technology firms are more focused on maximising user engagement to facilitate targeted advertising than considering the broader implications of their technologies.
Against this backdrop, a new paper from Frey et al. (2024) seeks to investigate alternative trajectories for AI development that pivot away from data-intensive methods towards approaches that require less personal data. The authors inherently frame their analysis within existing research on directed technological change, which argues that innovation tends to gravitate away from factors of production that are either scarce or costly. Cited examples include times of high wages prompting investment in automation and carbon taxes spurring green technology development.
Frey and his colleagues posit that regulations, specifically the General Data Protection Regulation (GDPR), have significantly increased the costs associated with storing and processing personal data, subsequently encouraging firms to invest in data-efficient methodologies instead. Their research reveals notable shifts in patent trends relating to AI technologies globally. They found that between 2000 and 2021, data-intensive AI patents surged at an annual growth rate of 52%, in contrast to the more restrained 19% growth of data-saving AI patents. The period following the GDPR’s implementation in 2018 marked a resurgence of interest in data-saving methods, with patents for transfer learning and synthetic data generation increasing by 185% and 86% respectively.
Moreover, the research underscores revealing geographic disparities in AI patenting. China's government strategy, known as "Made in China 2025," has fostered a significant expansion in AI patenting, largely driven by university and state institutions. In stark contrast, US companies have led in the data-saving category, contributing to 45% of relevant patents globally. The EU, while lagging in overall AI patenting, maintains a distinct advantage in indigenous data-saving innovations, pointing towards regional regulatory influences on technological development.
Markedly, the GDPR's effects aren’t limited to European firms; its stipulations impact global patent applicants targeting EU consumers. Analysis shows that companies directly affected by GDPR regulations have shifted focus towards less data-intensive AI approaches. This dynamic is especially pronounced among older and larger companies based in the EU. While this redirection has altered the technological landscape of AI, it has also contributed to a decline in overall patenting activity within the EU, consolidating market dominance among established firms.
Frey et al. conclude that the historical context of AI innovation has heavily emphasised data-intensive deep learning, leading to a relative neglect of less data-reliant methods. They draw parallels with the evolution of electric vehicles, arguing that focusing solely on one technology can lead to inefficient market dynamics. Emerging regulatory frameworks may further define this trajectory; the forthcoming EU AI Act, with its additional compliance burdens, could exacerbate market concentration and challenge smaller enterprises.
This analysis elucidates how policy interventions can deliberately steer the course of AI innovation, raising pertinent discussions about the long-term implications for the technological landscape in both Europe and globally. The intersection of privacy legislation and AI development continues to evolve, signalling a transformative period for businesses and researchers alike as they navigate this complex environment influenced by regulatory frameworks and emerging technologies.
Source: Noah Wire Services