While pursuing his PhD at MIT’s Center for Collective Intelligence, Erik Duhaime found inspiration in an unlikely place – his wife’s medical study routine. She was using educational apps loaded with quizzes and flashcards to prepare for her medical exams, which got Duhaime thinking. His research had already demonstrated that a group of medical students could collectively diagnose skin conditions more accurately than established dermatologists. The key was continually assessing each student’s proficiency, discarding those with less accuracy, and pooling the knowledge of the more proficient ones.
Thus, Duhaime conceived Centaur Labs and its flagship mobile app, DiagnosUs, in a bid to harness collective medical expertise and contribute to AI development. The platform allows users to evaluate real-world biomedical data, like images of potential skin cancers or heart and lung audio files, awarding cash prizes for accuracy. It bridges the gap between medical professionals’ aspiration to refine their skills and the pressing need for accurately labelled medical data by AI-driven biotech companies, pharmaceutical firms, and medical device developers.
Duhaime’s vision was driven by the concept of productive study for AI developers. “Why couldn’t studying for medical exams be a form of contributing to the AI community?” he thought. Today, the app attracts tens of thousands of users, with around 50% being medical students. Users are amazed that they can earn money while studying and refining their skills. The platform’s gamified approach adds a competitive edge, further enhancing the allure. This way, not only do users gain personal rewards, but they also aid in producing high-quality labeled data for AI projects focused on saving lives.
Duhaime’s PhD mentor was Thomas Malone, founding director of the Center for Collective Intelligence. Duhaime was drawn to the “wisdom of crowds” phenomenon, where collective answers tend to approach accuracy. His approach to utilizing this principle in complex tasks requiring expertise was to facilitate a platform offering ‘second opinions,’ amplified through the collective intelligence of many.
One experiment led him to train groups of lay people and medical students to classify skin conditions. He discovered that by pooling the top performers’ opinions, he could exceed the accuracy of professional dermatologists. This powerful approach combined AI algorithms with expert opinions, yielding results superior to either method individually.
Duhaime’s philosophy involves two core elements: measuring performance and identifying complementary expertise. Recognizing that different experts excel in different areas, he likens the process to assembling an optimal trivia team, where diverse skills and knowledge lead to a stronger performance overall.
Founded while Duhaime was still a student, Centaur Labs has become a hotspot for experts, including doctors, nurses, and medical students, who wish to test and enhance their skills. Not only does it provide a practical way to learn from real-life cases, but it also generates millions of expert opinions every week from users around the globe.
A major breakthrough came when Centaur Labs discovered that even medical students in Ghana could surpass experienced epileptologists in diagnosing EEG patterns. This proved that the platform could find and cultivate talent from unexpected quarters, furthering Duhaime’s vision of an interconnected, AI-assisted future where human judgment is seamlessly integrated with digital processes across various sectors of the economy. As Duhaime puts it, “It’s not just train algorithm, deploy algorithm,” but rather, a continuous feedback loop where humans and machines work in tandem, and platforms like Centaur Labs play a pivotal role in shaping this symbiotic future.
