AI Regulation Urged to Tackle Tech Monopolies, Bias and Job Displacement

Professor Shalom Lappin, affiliated with Queen Mary University of London, King’s College London, and the University of Gothenburg, argues that immediate policy interventions are required to mitigate the practical risks posed by the rapid development of artificial intelligence. His research, detailed in Understanding the Artificial Intelligence Revolution, highlights the concentration of AI development within a small number of large technology companies – 32 models in 2022 compared to three from universities – and the associated implications for research priorities and public benefit. Lappin’s analysis extends to the substantial environmental costs of AI, including the energy consumption of training large language models – approximately 50 gigawatt hours for ChatGPT-4 – and the resource-intensive manufacturing of necessary microchips. He advocates for international regulation of tech companies, reform of intellectual property rights to ensure fair compensation for data used in AI training, and proactive measures to address bias in AI systems, combat disinformation, and prepare for potential workforce disruption due to automation.

AI Development and Tech Monopolisation

Currently, a small number of powerful technology companies dominate the AI landscape, raising concerns about market concentration, innovation stifling, and equitable access to these transformative technologies. These companies control vast datasets, computational resources, and talent pools, creating substantial barriers to entry for smaller players and potentially hindering the development of a diverse and competitive AI ecosystem. Addressing this imbalance requires strategic policies that promote competition, foster innovation, and ensure that the benefits of AI are widely shared. The rapid advancement of artificial intelligence presents both unprecedented opportunities and significant challenges, demanding careful consideration of its societal impact and proactive policy interventions.

Their priorities often align with maximizing profits and reinforcing existing business models, potentially neglecting areas with significant societal benefit but limited commercial viability. This can lead to a skewed innovation landscape, where AI is primarily deployed in applications that serve narrow interests rather than addressing pressing global challenges like climate change, healthcare disparities, or educational access. Policymakers must actively encourage research and development in these critical areas through funding, incentives, and public-private partnerships, ensuring that AI serves the broader public good. Furthermore, fostering open-source AI initiatives and data sharing platforms can democratize access to these technologies, empowering researchers, entrepreneurs, and citizens to participate in the AI revolution.

Combating Bias and Disinformation with AI

AI decision-making systems demonstrably exhibit bias across critical sectors including healthcare, hiring practices, and financial services, leading to discriminatory outcomes and perpetuating existing inequalities. This bias often stems from flawed data, biased training methods, or inherent limitations in the algorithms themselves, demanding systematic correction and ongoing monitoring. Policymakers must establish clear standards for algorithmic fairness, requiring developers to assess and mitigate bias throughout the AI lifecycle, and holding them accountable for discriminatory outcomes. The proliferation of biased algorithms and the spread of disinformation pose serious threats to fairness, trust, and democratic processes, necessitating robust policy interventions and technological solutions.

The rapid dissemination of deliberately false information is amplified by the speed and scale of modern communication networks, eroding trust and undermining democratic processes. Obtaining consent from copyright holders before incorporating their protected material into training datasets is a minimum requirement, and new licensing models should be explored to facilitate fair compensation. Transparency regarding training data is crucial, obligating AI developers to publicly list the sources used to train their models, and international cooperation is vital, harmonizing intellectual property regulations or establishing agreements governing AI training data.