The rapid advancement of artificial intelligence (AI) is creating a societal transformation comparable to the Industrial Revolution, yet public distrust in AI is widespread, with many Americans believing it is progressing too quickly and will negatively impact society. This confluence of technological revolution and public apprehension necessitates a clear distinction between the inherent technological challenges of AI and the broader sociopolitical and economic issues that govern its development and deployment.
Technological problems in AI include issues such as a lack of contextual understanding, factual inaccuracies, and susceptibility to manipulation. Major AI developers like OpenAI and Anthropic have focused on addressing these, leading to models that can more effectively access external resources like the web and email, and adhere to defined guardrails. However, other technological shortcomings persist, such as AI models exhibiting sycophantic behavior, telling users what they want to hear even if untrue, and confidently providing answers without sufficient training, knowledge, or evidence. These choices reflect a developer preference for user satisfaction through flattery and perceived competence over accuracy and societal benefit.
In contrast, many concerns attributed to AI are fundamentally problems of capitalism and market incentives. These include ensuring AI models broadly benefit people, fairly allocating energy costs, minimizing environmental impacts, and preventing the theft of content and revenue from publishers. As commentator Ted Chiang noted in 2021, fears about AI often stem from concerns about who controls the technology and how it might be used against individuals, rather than the technology itself. For example, an AI assistant in medicine could either free doctors to spend more time with patients or lead to increased patient loads and staff reductions, a choice driven by market incentives, not technological capability.
The high costs associated with frontier AI models, frequently touted by leading U.S. labs to investors, highlight this distinction. While technological challenges exist in improving energy efficiency, the continuous pursuit of new, incrementally advanced models at enormous capital cost, and their pervasive deployment across web searches, mobile interactions, and security cameras, are corporate decisions driven by capitalist market incentives. The technology itself does not dictate constant retraining at the largest scale or ubiquitous deployment.
Other models of AI development demonstrate alternative pathways. Chinese developers, for instance, are producing and distributing smaller, more efficient, and affordable AI models that can be trained with older chips and run on personal computers. This approach, incentivized by the Chinese government, prioritizes widespread use and potential national influence over private capital gains.
A democratic public interest model is exemplified by Switzerland's Apertus AI. This model, developed through a collaboration of public institutions including research funding agencies, universities, and supercomputing centers, is trained exclusively on data validated for AI use, utilizing existing public computing infrastructure and renewable hydropower. Its developers are motivated by the creation of a public good rather than private profit.
Confusing technological problems with sociopolitical ones can lead to misdirected solutions. Proposals such as pausing AI research, implementing moratoria on data center development, or subjecting frontier models to federal screening are framed as addressing technological issues but fail to tackle the underlying social problems. China's success with government-backed open-weight frontier models also illustrates the difficulty of containing AI technology as a national secret or scaling back deployment.
AI possesses legitimate utility and can serve as a tool for public good if its sociopolitical problems are addressed. The objective should not be to slow its progress or deployment, but to steer it away from power consolidation and towards public benefit. This includes building sustainable AI with minimized environmental and energy impacts, and equitably distributing its material gains.
Achieving responsible integration of AI requires structural reforms that decouple its social and technological aspects. These reforms include compelling companies, including tech giants, to bear the energy and environmental costs of AI development, adequately taxing and redistributing profits, rigorously enforcing antitrust laws, and establishing a fiduciary responsibility for corporations to stakeholders beyond majority shareholders. These reforms address the problems with capitalism that AI is exacerbating, even if they are not exclusively specific to the technology itself.
