A robot made an appearance on stage while Nvidia CEO Jensen Huang was delivering a keynote address at the Nvidia GTC Artificial Intelligence Conference at SAP Center in San Jose, California on March 18, 2024. The image was captured by Justin Sullivan/Getty Images.
Kean Birch holds the position of director at the Institute for Technoscience & Society at York University.
The growing prominence of generative AI is being called into question in terms of its value in terms of time and financial resources. While technologies like ChatGPT offer some entertaining applications, they also bring about significant social drawbacks with limited societal advantages.
To grasp these drawbacks, it is essential to comprehend generative AI. It does not function as an autonomous, intelligent entity capable of reasoning and decision-making like humans. Rather, as highlighted by Emily Bender and her colleagues, generative AI replicates human actions by echoing our words and images. It operates by making guesses rather than engaging in actual thought, often resulting in what is referred to as AI hallucination.
Viewing generative AI as probabilistic systems sheds light on the social consequences that arise from their advancement.
AI relies heavily on computational power: the more AI systems we implement, the greater the demand for computing resources. This not only diverts computing resources from potentially more beneficial endeavors but also necessitates a substantial amount of energy. While the environmental impacts are well-documented, they are projected to worsen with the proliferation of AI. Sam Altman, the CEO of OpenAI, underscores the need for an “energy revolution” for AI to succeed. Apart from the ecological implications, the energy-intensive nature of AI will lead to increased energy costs throughout society.
Furthermore, AI is heavily reliant on significant capital investments in computing infrastructure. The foundation of AI rests on components such as fiber optics, servers, and data centers. The costs associated with this infrastructure are evident in the financial reports of Big Tech companies, which highlight the billions allocated and being spent on such infrastructure. Although Big Tech now controls a substantial portion of our computing resources (which in itself is a social cost), additional investments are required to make AI commercially viable as an everyday technology. These investments could potentially be directed towards more productive purposes.
AI is also absorbing a considerable amount of innovation funding, particularly from venture capital. According to CBS Insights, venture capital expenditure on generative AI surged fivefold between 2022 and 2023, reaching nearly US$22 billion. In a shrinking pool of venture capital, these funds could have been allocated elsewhere. Moreover, commentators like Ed Zitron point out that if the AI hype bubble bursts, as seems probable, all the innovation funding and capital investments will have been squandered.
As AI progresses along this path, it runs the risk of inundating us with AI-generated content. AI relies on data to train its models, but content creators such as newspapers, websites, and authors are pushing back against the unauthorized use of their copyrighted material by litigating against organizations like OpenAI. More critically, as AI saturates the internet with AI-generated “data,” it may collapse under its own weight. This scenario, as described by political economist Jathan Sadowski, paints a picture of “Habsburg AI,” where artificial intelligence technologies become so interbred with outputs from other generative AIs that they mutate into grotesque forms with exaggerated features. This leads to a cascade of hallucinations, resulting in a myriad of unforeseen repercussions.
Perhaps the most crucial point is that AI shifts the burden of its social impacts onto the broader society, even when it fails to provide any substantial social benefits. The integration of AI will inevitably bring about significant societal changes and associated costs as we are compelled to reshape our social, political, and economic structures to address the repercussions of its effects. Even seemingly basic applications like AI-generated images will incur collective costs when it comes to mitigating their influence on our political institutions. For instance, adapting our political system to combat the surge of political misinformation amplified by generative AI will require substantial resources.
At the core of the issue lies the fact that generative AI is not inherently geared towards addressing real societal challenges. There is an urgent need for the insights of social scientists to guide collective decision-making regarding the future trajectory of generative AI. This responsibility cannot be left solely to businesses, markets, or technologists. It is imperative to engage these experts in identifying our social or collective issues and determining the challenges that generative AI should tackle. Subsequently, the focus should shift towards exploring whether—rather than just how—artificial intelligence can contribute to devising viable solutions, prompting AI companies to concentrate on developing these solutions.
