A Call to Action on AI, From Inside Silicon Valley
We spoke with some of the academics who co-signed Stanford Critical AI's Manyfesto. The document, and its signers, say "outside this campus, something is happening that is too loud to ignore."
An interdisciplinary group of academics from around the world have signed their names to a “Manyfesto” (a name that nods to the AI Decolonial Manyfesto) out of Stanford University’s Critical AI group.
The opening statement is succinct and digestible, and I encourage you to take a moment to read it, but its six points of belief in summary are:
AI is a socio-political technology.
AI should be governed as a normal technology.
AI’s trajectory is not predetermined.
To be critical of AI is to be reasonable.
To be critical of AI is to question power and information asymmetries.
The most important AI harms are currently hidden.
The rapid and unregulated development of AI, the dominant narrative of its inevitability, and its effects on labor, environment, war, and government surveillance — civilization, as one co-signer summarized — have converged in an urgent way this year. From a university famous for its status in Silicon Valley, an open call to action on AI is significant. Its creators and co-signers are highly aware of this. They say Stanford’s longtime influence in technological development comes with responsibility to criticize the trajectory and consequences of those technologies.
I got to speak with four co-signers and one co-founder of the Manyfesto, all from different research backgrounds and Universities. Here’s what they said.
Conversations edited for length and clarity.
Chijioke Mgbahurike, a Law & AI Research Fellow with Stanford’s Liftlab, is one of seven co-founders of Stanford Critical AI, which officially launched in March. His research focuses on sociotechnical implications of AI bias on marginalized communities. He told me the project has been in development since last October, when he and other co-founders attended a speaker panel about AI and warfare, and came away from it frustrated.
“I think the main impetus was the feeling that Stanford was not very critical towards AI in ways that we found satisfactory, and that rather than waiting for someone to change that culture, that we can just change it ourselves,” he told me. “I know some popular figures think about [criticism] like AI populism, or like the backlash towards data centers, as this group of people who are under-informed… I’ve seen a lot of these narratives come through both through Stanford and the Bay more generally… When I was at that particular panel, I could appreciate the analysis, but I realized in that moment there was nothing new that I didn’t already know. And I thought that I was kind of looking for action as opposed to some kind of analysis.”
As for why the group decided on a manifesto, Mgbahurike said the political nature of this kind of document rises to the political nature of AI as a technology, and of Stanford’s place in its ecosystem.
“We must recognize that, with the sort of spread of tech manifestos, what we’re seeing is this combination of both politics and technology,” he said. “We need to, as a community, take a look at what our institutions created, and what we ourselves want to create in that world.”
Co-signer Dr. Lucy Suchman is an anthropologist who has had a critical engagement with artificial intelligence for over 40 years. Her research has focused on, among others, the politics of technology design and militarization of AI. She told me there’s a direct throughline from concerns in the 1980s about the automation of control systems for nuclear weapons to the presence and use of AI in militaries today.
“I’ve always been interested in the social, political, economic relations that give rise to these technologies and that inform the way that they’re developed. And so now we’re seeing those political economies in particular reaching their sort of apex,” she said. “They’re sort of playing out to their logical extensions as venture capitalist speculative investments.
“Consistently, those who are most prominent in relations, particularly to government inquiries into AI and its regulation and its problems, are the same actors. We have the military commercial complex and the revolving door where people are constantly moving between government and private sector positions. The idea is, well, we’re talking about these technologies. We have to get the people with the most expertise. It turns out that they’re the same people who are running these companies. And so they end up on the boards that are advising the government. They end up effectively writing the legislation, having enormous influence. And this absolutely marginalizes, and in many cases, kind of shuts down other voices.”
Dr. Wenshan Jia is a communications scholar who researches the impact of technology on international intercultural communication. He told me studying AI through that lens is incredibly unique to traditional media.
“This is something which tries to emulate, even replace human agency,” he said. “It’s very unique in that ChatGPT or large language models were developed without the input of a peer review process… The development of the large language models itself was developed with the money from the Corporations, and it’s really close-ended, no peer review process. So therefore, when they developed it, there was nothing in the process of making sure that they’re ethical. And so as a result of that, they create lots of problems. And most importantly, technologically speaking, we are at this narrow intelligence period, but the next stage is artificial general intelligence, and the third stage is artificial super intelligence. And most importantly, humanity will be enslaved.”
He’s concerned about human agency in the age of AI, and the survival of critical thinking in young people who come through school using generative language models to produce work for them — but as the Manyfesto itself concurs, he is not against the use of AI overall, even within academia. As a communications researcher, he says that developing a media literacy framework specific to AI is crucial.
“AI literacy is quite different than social media literacy, than internet literacy, or media literacy in general,” he said. “I think [the general public] need[s] to be informed of the potential consequences of the massive use of AI in all aspects of life. They need to be providing guardrails, they need to be given options when to use it, how to use it. If AI is present, they’d be aware of it and how to reduce it, how to develop a kind of a healthy relationship between AI and them. And that’s why I think AI literacy is very significant. It should be a kind of education which cuts across all society sectors.”
Dr. Andrés Burbano studies the history of media technologies, and is involved in a historical initiative to track the evolution of the critique of artificial intelligence. He told me the impact on distribution of political information in the age of AI and social media, and its effect on democracies everywhere, should be one of the priorities of our times.
“These transformations are affecting all of us, every person with a mobile phone in their hands. But even beyond that, people that don’t have access to communication systems are affected by these transformations,” he said. “We need to have more information about the history of how the field took shape as a technology, AI, as a set of technologies. But also to identify that critical thinking and a critique of this kind of technology has been present for a long while. That is not something that the present brings.”
Dominic Zappia, a research associate with Stanford’s Technology Ethics & Policy Rising Scholars Program with a background in economics, philosophy, and political science, told me he sees how AI is regulated and who regulates it as a reflection of the wider sociopolitical moment.
“I think that it’s furthering this general sort of sense that people have that they’re not in control of what’s going on. The feeling of being left behind embodied in the populist movements of Trump,” he said. “I would argue that people feel very little positive benefits from AI in their daily life. And if they are positive benefits, they’re diffuse and the harms are very concentrated. The harms of AI diffusion are the data centers. They’re people losing jobs. So when you see this thing happening and you feel like it’s bad, and there’s nothing you can do about it, that continues to erode trust… And there’s also general resentment against Big Tech. It’s bad, is the takeaway. And I think that Silicon Valley doesn’t feel that as much as it should.”
The Manyfesto is a call for open criticism of AI as a method of intervention. Its interdisciplinary backers highlight, through the lens of their different areas of research, that AI as a technology, as well as the pace of its development and exponential prominence in everyday life, is everyone’s problem. Questioning AI’s inevitability and its presence in any given space is important, and not to be dismissed as “technophobic” or anti-progress. Everyone I spoke to emphasized that the general public, as powerless as we can feel over AI’s takeover, retains this ability and responsibility to question. Academia, and the Manyfesto, is simply a framework for criticism.
“My hope has been that when people find out about our group, a lot of people are surprised that this could come out of Stanford,” Mgbahurike told me. “One of the goals of the work is that, ‘you guys are not alone, you guys are not powerless’. There are people who are fighting for the basic dignities that we believe you are entitled to. And that just because you cannot make it into kind of these, quote unquote—I use this as a derogatory term—hallowed spaces of Stanford or any of the higher level institutions does not mean that your ideas, your beliefs and your wants no longer matter.”
Stanford Critical AI says the project will aim to build community, advance AI research, and create spaces for advocacy on campus and beyond, to reiterate AI criticism as a contemporary civic and academic responsibility. The Manyfesto and its co-signers want you to know, no matter what is framed as inevitable: the future of AI is still a matter of public choice, and the public is entitled to questioning it.