17-year-old wins $250,000 science prize for making AI safer and more efficient
-
Achyuta Rajaram, a 17-year-old high school student, won the $250,000 top prize at this year's Regeneron Science Talent Search for his project on making machine learning more efficient and safer.
-
The competition awarded over $1 million in total prizes this year and received over 2,000 applications, the highest number since the 1960s.
-
Rajaram created an algorithm to automatically find and analyze specific components of neural networks called circuits, which could help explain how models reach conclusions.
-
He also demonstrated how to modify a text-reading neural network to fix a vulnerability where adding text could cause it to misclassify images.
-
Rajaram believes understanding neural networks better will allow computer vision and AI systems to become faster, more accurate, and safer overall.