In mid-2021, a canvassing of 602 technologists, policy leaders, researchers, and activists by Pew Research Center and Elon University’s Imagining the Internet Center revealed a stark divide over the trajectory of ethical artificial intelligence. When asked whether, by 2030, most AI systems would employ principles focused primarily on the public good, 68% said no. Only 32% expected—or at least hoped—that such principles would guide the majority of systems.
The concerns voiced were wide-ranging. Many argued that defining “ethical” AI is inherently difficult. Cultural differences, shifting norms, and the complexity of real-world contexts make universal standards elusive. Stephen Downes noted, “The problem with the application of ethical principles to artificial intelligence is that there is no common agreement about what those are.” Others pointed to the “black box” nature of modern machine learning, where even creators may not fully understand decision pathways, complicating oversight.
A recurring theme was the concentration of AI development in the hands of powerful corporations and governments, often driven by profit or geopolitical advantage rather than moral imperatives. Amy Webb warned of China’s expansion of surveillance technologies and the strategic vulnerabilities posed by “rogue actors” capable of disrupting infrastructure. Stowe Boyd cautioned that corporate control could channel AI toward wealth concentration rather than broad societal benefit.
Several experts highlighted that AI often inherits and amplifies human biases embedded in training data. danah boyd observed that most large corporations “fetishize efficiency, scale and automation” in ways that entrench inequities, arguing for a focus on augmentation and inclusion instead. Marcel Fafchamps emphasized that while AI integrates ethical principles by design, “the fact that AI integrates ethical principles does not mean that it integrates ‘your’ preferred ethical principles.”
Some respondents saw glimmers of hope. Ethan Zuckerman credited activists for putting AI ethics on the agenda, creating “a rare opportunity to deploy AI in a vastly more thoughtful way.” Ben Shneiderman suggested adopting aviation-style “flight data recorders” for AI systems to enable forensic analysis and enforceable improvements. Marjory S. Blumenthal anticipated credible advances in medical diagnosis and personalized education if privacy and safety are prioritized.
Others were more skeptical about systemic change. Susan Crawford bluntly stated, “We have no institutions that could impose those constraints externally.” Mike Godwin predicted that policy responses would be reactive and piecemeal, triggered by publicized abuses. Sam S. Adams argued that universal access to AI tools makes it “practically no way to force ethical use in the fundamentally unethical fractions of global society.”
The potential for misuse was a constant undercurrent. Jonathan Grudin foresaw AI’s principal use remaining in “finding ever more sophisticated ways to convince people to buy things that they don’t really need.” Calton Pu warned of “bad” AI applications exploiting the gap between static training data and evolving reality, a vulnerability already demonstrated by Microsoft’s Tay chatbot. Dan S. Wallach pointed to military autonomy as a domain where ethical prohibitions are unlikely to hold.
Even optimistic projections acknowledged challenges. Vint Cerf expressed skepticism that good intentions would reliably yield desired outcomes, citing the unexplored design space of machine learning. Susan Etlinger likened the evolution of AI norms to the gradual adoption of safety measures in the auto industry, but cautioned that “all bets are off” with AI-enabled weaponry. Esther Dyson hoped for greater transparency in both algorithmic and human decision-making, contingent on society’s willingness to act on uncomfortable truths.
For engineers and technologists, the discussion underscored that AI’s impact will be shaped as much by governance, incentives, and cultural values as by technical capability. Barry Chudakov urged a reimagining of ethics itself in the age of big data, where “you create the answer by the question.” Beth Noveck called for participatory design processes involving those most affected by AI tools, to prevent perpetuation of bias.
By 2030, AI will almost certainly be more capable, more pervasive, and more embedded in daily life—from autonomous vehicles and healthcare diagnostics to logistics optimization and human–machine interaction. Whether these systems will primarily serve the public good remains, in the eyes of most surveyed experts, unlikely without significant shifts in policy, education, and industry practice. As David Karger put it, aligning corporate profit motives with societal benefit will require deliberate public action and enforceable laws.