Women in AI and Government at Cornell: Sarah Kreps
In the male-dominated tech industry, it’s important to shine a spotlight on the remarkable women who have contributed to the AI revolution. TechCrunch is launching a series of interviews to give these women the recognition they deserve. One such woman is Sarah Kreps, a political scientist, U.S. Air Force veteran, and analyst who focuses on U.S. foreign and defense policy. Kreps is currently a professor of government at Cornell University, an adjunct professor of law at Cornell Law School, and an adjunct scholar at West Point’s Modern War Institute.
Kreps’ recent research delves into the potential and risks of AI technology, particularly in the political sphere. In an opinion column for The Guardian, she discussed the intensification of the AI arms race across companies and countries as more money pours into AI. She also highlighted the challenges surrounding AI policy.
Sarah Kreps’ journey into the field of AI began with her background in emerging technologies with national security implications. As an Air Force officer, she was involved in advanced radar and satellite systems. This experience naturally led her to study the national security implications of emerging technologies as a PhD candidate. Initially, Kreps focused on drones and their autonomy, which brought her into the realm of artificial intelligence.
It was during an artificial intelligence workshop in 2018 that Kreps first encountered OpenAI’s GPT-2 capability. As she listened to the presentation, she became preoccupied with the potential of generating credible disinformation at scale and manipulating the psychology of American voters through microtargeting. Intrigued by these possibilities, Kreps reached out to OpenAI and became one of their early academic collaborators. Her research aimed to investigate the credibility of GPT-2 and later GPT-3 as political content generators. Through a series of experiments, she evaluated whether the public would perceive this content as credible and conducted a large field experiment to determine if legislators could be fooled by a large-scale letter writing campaign.
Kreps’ work in the AI field is multifaceted, but she is particularly proud of the field experiment she conducted. This experiment demonstrated the disruptive potential of AI in shaping the legislative agenda. Additionally, she collaborated with computer science students at Cornell to develop an application that would assist legislators in responding to constituents’ emails. While these tools were never brought to market, they explored the intersection of AI and politics, addressing constituents’ disaffection and legislators’ time constraints.
Navigating the male-dominated tech industry has not been a significant challenge for Kreps as a researcher. However, she acknowledges the intimidating atmosphere of events dominated by men. Her advice to women entering the AI field is to find mentors, develop skills, let their abilities speak for themselves, and remain resilient in the face of challenges.
When it comes to the pressing issues facing AI, Kreps raises concerns about aligning AI with certain values without considering the broader societal implications. She cites the problematic rollout of Google Gemini as an example, where aligning with a narrow set of developers’ values led to historical inaccuracies in the outputs. These large language models shape how people think about politics and social relationships and confer considerable power onto the big firms responsible for developing them.
AI users should be aware of the need for verification in a world saturated with AI-generated content. Kreps advises users to be circumspect and seek alternative sources to verify authenticity before unquestioningly trusting AI-generated information. This lesson has already been learned with social media and misinformation.
In terms of responsibly building AI, Kreps emphasizes the importance of scientists being public stewards. In a piece for the Bulletin of the Atomic Scientists, she outlines steps for responsible development while challenging some of the questions posed by AI developers as being incomplete or misguided.
Sarah Kreps’ contributions to the field of AI and her exploration of its implications in government and politics are invaluable. As the AI revolution continues, it is crucial to recognize and celebrate the achievements of women like Kreps who are shaping the future of AI.