Introduction

Contemporary AI development is accompanied by widespread industry-led hype: speculative narratives that portray AI as transformative, inevitable, and urgently necessary. These narratives contain sweeping claims about AI’s superhuman capabilities, capacity to drive economic growth, and potential to solve complex societal challenges. The production of these narratives has intensified since the 2022 release of ChatGPT. This proliferation occurs alongside a rapid entrenchment of AI across sectors, where firms have accrued immense political and economic power, with valuations skyrocketing to unprecedented heights.

AI leaders regularly convene with politicians, and governments have passed AI-friendly policies, removing regulatory barriers while facilitating the accelerated construction of data centres. Meanwhile, AI has permeated society, with institutions incorporating systems into core operations. This has concentrated control in a handful of unaccountable firms while producing harms such as environmental degradation and exploitative labour conditions, with severe consequences for marginalised groups.

The idealist account of AI hype

The dominant account of AI hype frames it as a problem of misrepresentation, referred to as an “idealist” account. This perspective challenges the accuracy of claims, labelling many AI systems that cannot work as advertised as “snake oil”. Accordingly, this account suggests challenging hype by promoting technical and sociotechnical literacy. However, the authors argue that the idealist account only minimally connects hype to the AI industry’s growing power and lacks a theory of change for altering that power structure. Literacy efforts often concede to the narrative of inevitability and can even be invoked by hypers to deflect criticism.

The purposes and impacts of AI hype

A materialist deconstruction defines tech hype as a performative, promissory discourse that mobilises visions of technological futures to organise action in the present. Hype serves as a discursive strategy to coordinate audiences around actions crucial to the economic success of emerging technologies, such as marshalling resources and providing investments. Hypers pursue two goals: capturing credibility (making speculative futures appear believable and desirable) and manufacturing urgency (framing immediate action as necessary and irrefutable).

Defining tech hype

Hype is an industry-agnostic mechanism for circulating value and expectation within contemporary capitalism. It is productive, bringing desired futures into existence by bolstering emerging markets, defining research agendas, and justifying infrastructure development. Hype is considered “successful” if it shapes policies and markets in ways that benefit the hypers, regardless of the presence of inaccurate or exaggerated claims.

Situating AI hype

AI hype is the manifestation of tech hype under the conditions of contemporary AI development. The industry is heavily dependent on hype because its ventures are cost-intensive and uncertain, requiring enormous capital for infrastructure like data centres. The paper distinguishes between primary hypers (tech CEOs, venture capitalists) whose core interests are tied to AI entrenchment, and secondary hypers (policymakers, corporate managers) who reproduce hype to serve external institutional interests, such as attracting tax revenue or appearing innovative.

How AI hypers build and sustain hype

Hypers use a discursive playbook to capture credibility and manufacture urgency. This involves distorting empirical knowledge by suppressing limits, aligning actors around shared priorities by embedding technologies within shared imaginaries, and delimiting trajectories of development by foreclosing alternatives even in the face of criticism.

Capture credibility

Hypers build support by making themselves and their speculative futures appear credible. Discursive practices include enchanting audiences through “spectacle” and technosolutionist narratives, such as framing AI as a “general-purpose solution” to diverse societal challenges. They also mobilise dominant imaginaries, such as geopolitical competition (the “AI race” against China), to shore up support. Furthermore, they adapt claims to deflect criticism, reframing harms as problems manageable through better governance or technical expertise.

Manufacture urgency

To sustain material investment, hypers position AI development as urgent and difficult to refuse. Tactics include mischaracterising the past by obscuring historical precedents of failure and cultivating a “fear of falling behind” to encourage rapid adoption. They present the future as inevitable, using deterministic rhetoric to foreclose debates. For example, industry leaders portray AI adoption not as a strategic choice but as a necessary condition for survival, forcing institutions to adapt rather than deliberate.

A playbook for resisting AI hype and entrenchment

Resistance requires contestation rather than just literacy. Contestation involves building counterpower among the groups on whom AI is imposed, such as workers and communities. The goals are to undermine hypers’ legitimacy and build collective political agency. This approach pursues “non-reformist reforms” that seek to redistribute power and build capacity for democratic control across society.

Undermine hypers’ legitimacy

Opponents should dislodge hypers’ position as credible stewards. Tactics include disenchanting audiences using counter-spectacle, such as the defacing of New York City subway ads for the AI startup “Friend” with messages like “Stop normalising this”. Critical research, such as the Data Workers’ Inquiry, also disenchants audiences by exposing the exploitative labour conditions that enchantment conceals. Opponents also advance alternative imaginaries and call out self-serving rearticulations, such as deconstructing corporate claims about “AI climate solutions”.

Build collective political agency

Opponents must grow collective power to make choices about AI’s trajectory. This involves recovering suppressed histories of AI’s repeated failures and resistance, such as the 19th-century Luddite movement. Opponents should create material friction to AI expansion through protests, boycotts, and ballot initiatives. Cross-partisan organising has already blocked projects worth several hundred billion dollars. Finally, expanding political imagination involves reopening questions about whether AI expansion should proceed and demanding collective control over infrastructure.

Conclusion

Taking AI hype seriously as a material force reveals a more effective roadmap for resistance. Resisting hype requires disorganising institutional alignments and building collective power. While AI hypers currently hold significant power, movements must aggregate local victories and sustain pressure to shift entrenched arrangements. Contesting hype is essential for wresting power from hypers in order to shape technological futures themselves.