It’s not the speed, it’s the direction

After refusing to join hands earlier this year, it seems that the leaders of the world’s biggest artificial intelligence (AI) research labs have found something they can agree on. Last month, the CEOs of Anthropic, OpenAI, DeepMind, Microsoft, and SpaceXAI (previously xAI) have called for “pacing” AI research. Many people, including these leaders, claim that AI research poses significant risks. This is because these companies are not merely trying to build helpful products or tools to improve people’s lives. Rather, the stated goal of their research is advanced “artificial general intelligence,” defined by OpenAI as “highly autonomous systems that outperform humans at most economically valuable work.” The “pacing” that they call for refers to the speed at which AI companies pursue this goal.

But we should not be misled: the pace is not the issue. It doesn’t matter how slowly they do it if what AI research labs are building is fundamentally something the public does not want. “Pacing” presumes that the course these companies are taking us down is inevitable. But we don’t just need a different pace. We need a better direction.

While the rate of public adoption of generative AI systems is high, concerns are growing. Last month, Pew Research Center shared findings from a global survey on attitudes toward AI, showing that people in many countries across different economic and cultural contexts expect AI to lead to fewer jobs and greater inequality. (This sentiment is shared by many AI builders, themselves. Anthropic’s Dario Amodei previously projected the elimination of half of all white collar work by 2030.) In the United States, the growing protests against Flock cameras and data centers represent to many not just a debate about physical infrastructure, but growing public anger about big tech, AI, and surveillance. 

These frustrations are not confined to the pace of development. Instead, they reflect the disconnect between the priorities of AI builders and those of the public. Last year, Pew Research Center found that 59% of Americans are not confident in companies developing and using AI responsibly. Yougov surveys report that the spread of deep fakes, erosion of personal privacy, manipulation of human behavior, and increased human dependency on technology are top concerns. Building systems that “outperform humans at most economically valuable work” may be the end goal of many AI researchers, but it isn’t what most people want.

This doesn’t mean that people are resistant to technology in general. In my work as AI, Labor, and Economy program lead at Partnership on AI and more recently as Director of Emerging Technologies at Aspen Digital, I have worked with many communities on the frontlines of AI-related labor disruption. Far from the stereotypes maligning them as technology-rejecting “luddites,” these communities often hold nuanced perspectives on AI. Members of many labor unions, for instance, support technologies in the workplace that improve workers’ safety, agency, and skill-building, but they push back on worker surveillance and systems that automate important decisions about performance, hiring, firing, and compensation. 

Many labor advocates have pushed for worker voices to be incorporated earlier into the AI research and development process so that the technology better reflects their goals and priorities. Broadly, people want to have a say in the way that technology is built and how it shows up in their lives. They continue to see potential benefits from AI, but people want to know when it is being used, they want it to work for them, and they certainly don’t want AI to do all things. As the viral 2024 tweet from Joanna Maciejewska puts it: “I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.” 

While deep-pocketed developers in the AI industry rush to develop “general-purpose” AI systems, they are dragging their feet on important innovation that centers people. AI can be designed to promote widespread flourishing, but that is a design decision developers have to prioritize. Rather than slowing down, we need them to accelerate on transparency, promoting human autonomy (including protecting privacy, interpersonal connection, and cognitive agency), and making concrete progress on the most pressing issues of our time, like climate change and world hunger, rather than abstract notions of “intelligence.”

We do not need to give in to the AI industry’s latest attempts at avoiding regulation through voluntary so-called pacing. We should, instead, be requiring them to more quickly build capacity for public governance of these powerful technologies.

In fact, we can’t afford to wait.