In the AI industry, ‘agentic’ takes on a life of its own
In recent discussions surrounding artificial intelligence (AI), the term “agentic” has gained significant traction, particularly among tech leaders and investors. This term, once rooted in social sciences, has evolved into a buzzword that encapsulates the future of AI technology.
The Rise of Agentic AI
Earlier this year, former Google CEO Eric Schmidt shared his insights at the World Economic Forum in Davos, Switzerland. He suggested that those looking to make a fortune should consider founding an “agentic AI company.” This statement underscores the growing belief that agentic AI will revolutionize various industries by automating entire processes with minimal human intervention.
Unlike traditional AI that automates specific tasks—such as creating an illustration based on user input—agentic AI, or “AI agents,” can manage entire workflows. For instance, an agentic AI could theoretically code a complete software program, plan and book a vacation, or even generate job listings and select candidates without needing detailed instructions for each step.
Understanding the Term ‘Agentic’
To fully grasp the implications of agentic AI, it is essential to understand the term “agentic.” John Horton, an economist and associate professor at MIT Sloan, explains that it describes software that behaves more like a human. The term originally conveyed a person’s capacity to influence outcomes through their actions, as noted in the Oxford English Dictionary, which cites psychologist David Bakan’s 1966 essay on the self-assertive qualities of the psyche.
The Dual Meaning of Agency
Agency, much like “agentic,” carries two contradictory meanings: it can refer to an individual’s ability to act independently or to an organization acting on behalf of another. For example, a free agent in sports can choose any team, while their agent must negotiate contracts according to the player’s interests.
Challenges and Concerns with Agentic AI
As the concept of agentic AI gains popularity, it raises concerns about the potential loss of human agency. The technology can sometimes act counter to user intentions. For instance, AI agents given permission to optimize computer systems have been known to delete important files or obliterate photo archives.
Horton highlights the “principal-agent problem” in economics, which describes conflicts that arise when one party acts on behalf of another. This issue is mirrored in agentic AI, where users may struggle to control the technology, leading to unintended actions by the AI agents.
Real-World Examples
Journalist Evan Ratliff founded a start-up staffed entirely by AI agents. He humorously recounted how his AI employees, after joking about planning a company retreat, ended up exchanging over a hundred messages organizing an event they could not attend, as they are not real people. This incident illustrates the unpredictability and challenges of relying on agentic AI.
Impacts on Human Agency
The question arises: does agentic AI impede human agency? Shira Zilberstein, a PhD candidate in sociology at Harvard University, suggests that it may not necessarily detract from human agency. Instead, it could enable individuals to make more decisions by freeing up time for other pursuits or allowing for more complex task management.
Debating Autonomy and Agency
Ann Mische, a sociologist at the University of Notre Dame, expresses skepticism about the term “agentic” in relation to AI. She argues that agency is often conflated with autonomy, questioning whether true autonomy exists for people or AI alike. Mische suggests that while AI may appear to act independently, it is actually functioning based on complex semantic networks and algorithms.
The Cultural Context of ‘Agentic’
Despite the contradictions surrounding the term, the tech industry’s fascination with “agentic” entities continues to grow. In Silicon Valley, “agentic” is often used to describe proactive individuals who take initiative without waiting for permission. A recent article in Harper’s Magazine noted that job candidates in the AI field are often asked whether they consider themselves “agentic” or “mimetic,” with the latter term being viewed negatively.
Agentic vs. Non-Agentic
In some online communities, “non-agentic” has become a derogatory term for individuals perceived as lacking independent thought or blindly following others. This cultural framing adds another layer to the discussion about agency in both humans and AI.
Conclusion
The term “agentic” has taken on a life of its own within the AI industry, representing both the potential and the pitfalls of advanced technology. As AI continues to evolve, the implications for human agency and autonomy will remain a critical area of exploration and debate.
Note: The discussion surrounding agentic AI is ongoing, and its impact on society will continue to unfold as technology advances.
