In this recent article in SciTechDaily, writer Adam Zewe from the Massachusetts Institute of Technology (MIT) explains how thinking like a neuroscientist can help address dataset bias in AI. Zewe shares how a group of researchers from MIT, Harvard University, and Fujitsu Ltd., gained insights into how machine learning can overcome bias.
The group’s research was focused on neural networks, “a machine-learning model that mimics the human brain in the way it contains layers of interconnected nodes, or ‘neurons’ that process data.” The research identified that more diverse datasets enable the network to overcome bias.
“But it is not like more data diversity is always better; there is a tension here. When the neural network gets better at recognizing new things it hasn’t seen, then it will become harder for it to recognize things it has already seen,” states Xavier Boix, a research scientist in the Department of Brain and Cognitive Sciences (BCS) and the Center for Brains, Minds, and Machines (CBMM), and senior author of this research paper.