DiagnosUs: Revolutionizing AI with Cash Rewards and Human Expertise

DiagnosUs, an innovative mobile app, is transforming the field of artificial intelligence (AI) by harnessing the collective intelligence of medical students and professionals. By gamifying the process with cash prizes for accurate diagnoses, DiagnosUs enables medical AI companies to enhance their algorithms effectively. This groundbreaking approach, pioneered by Centaur Labs, consistently delivers reliable results that often match or surpass those of expert diagnoses. Erik Duhaime, the visionary behind DiagnosUs, envisions a future where his company continuously monitors AI models, creating an integrated ecosystem that combines human expertise with AI technology.

The platform developed by MIT alumnus Erik Duhaime taps into the wisdom of crowds to label medical data for AI companies. Centaur Labs has created an app that allows experts to classify medical data and receive small cash prizes in return. These expert opinions serve to train and improve life-saving AI models.

Duhaime’s inspiration for creating DiagnosUs stemmed from observing his wife, a medical student, spending hours studying on apps that offered flashcards and quizzes. His research revealed that, collectively, medical students could classify skin lesions more accurately than professional dermatologists. The key was to continuously assess each student’s performance on cases with known answers, disregarding opinions from individuals who performed poorly while aggregating the opinions of those who excelled.

By combining his wife’s study habits with his research findings, Duhaime founded Centaur Labs. The company introduced DiagnosUs, a mobile app that gathers the opinions of medical experts on real-world scientific and biomedical data. Through the app, users review a wide range of materials, including images of potentially cancerous skin lesions and audio clips of heart and lung sounds that may indicate health issues. If users provide accurate assessments, Centaur incorporates their opinions into its algorithms and rewards them with small cash prizes. This symbiotic relationship between users and medical AI companies leads to improved training and algorithm enhancement.

The Centaur Labs co-founders, Tom Gellatly, Erik Duhaime, and Zach Rausnitz, have created a gamified platform where individuals compete to train data while honing their skills and earning money simultaneously. The app has gained popularity, attracting tens of thousands of users worldwide. Remarkably, around half of the user base consists of medical students who are pleasantly surprised by the opportunity to earn money while studying. DiagnosUs offers an engaging platform that enables users to test their skills by reviewing actual medical cases, surpassing the traditional approach of multiple-choice questions often found in exam preparation.

Centaur Labs’ approach not only revolutionizes medical data labeling but also distinguishes itself from conventional practices such as outsourced data labeling and AI content moderation, which are typically assigned to low-resource countries.

The accuracy of Centaur’s approach has been demonstrated in various studies. Collaborating with researchers from renowned institutions such as Brigham and Women’s Hospital, Massachusetts General Hospital (MGH), and Eindhoven University of Technology, Centaur successfully showed that crowdsourced opinions labeled lung ultrasounds as reliably as experts did. Additionally, in partnership with researchers from Memorial Sloan Kettering, Centaur demonstrated that crowdsourced labeling of dermoscopic images surpassed the accuracy of highly experienced dermatologists. The scope of Centaur’s platform extends beyond images and includes video, audio, text from research papers or anonymized conversations between doctors and patients, as well as waves from electroencephalograms (EEGs) and electrocardiograms (ECGs).

Centaur’s journey has revealed that exceptional performers can emerge from unexpected sources. In a contest held through the DiagnosUs app at a conference featuring over 50 epileptologists with more than a decade of experience, it was two medical students from Ghana, Jeffery Danquah and Andrews Gyabaah, who outperformed all the attendees. This outcome highlights the powerful learning opportunities provided by DiagnosUs, as expressed by Gyabaah, who mentioned how the app’s insights improved his performance in real clinical settings.

Looking ahead, as AI continues to reshape the future of work, Duhaime believes that Centaur Labs will serve as an ongoing checkpoint for AI models. While the platform currently focuses on training algorithms, Duhaime envisions its role expanding to include algorithm monitoring and serving as a human-in-the-loop component for various tasks. Rather than viewing Centaur as solely a means to train AI, he sees it as an integral part of the entire life cycle, providing feedback on model outputs and monitoring their performance.

Duhaime predicts a future where human expertise and AI algorithms seamlessly integrate, and Centaur Labs is poised to play a vital role in this paradigm shift. As digital assembly lines become ubiquitous across industries, the infusion of on-demand expert human judgment at different stages of the value chain will be essential, replacing the traditional model of training algorithms and deploying them in isolation.

Frequently Asked Questions (FAQs) about AI revolution

More about AI revolution

• DiagnosUs App
• Centaur Labs
• MIT Center for Collective Intelligence
• MIT Sandbox Innovation Fund
• MIT Martin Trust Center for MIT Entrepreneurship
• Y Combinator
• Brigham and Women’s Hospital
• Massachusetts General Hospital (MGH)
• Eindhoven University of Technology
• Memorial Sloan Kettering