Funded by the IARPA MICrONS program over the past two years, Sebastian Seung’s lab atPrinceton University has built a system that enables humans to proofread an automated reconstruction generated by artificial intelligence (AI). The MICrONS system was used to reconstruct all connections between 191 neurons in a million cubic micron volume (terascale dataset) of mouse visual cortex, which required39 hours of proofreading by two graduate students. This massive speedup has been made possible by the ongoing “deep learning”revolution. Ten years ago, the Seung lab (then atMIT) first applied convolutional nets to EM images(Jain et al. 2007). Starting in 2012, corporations and governments poured billions of dollars into software and hardware for convolutional nets. The Seung lab has demonstrated “superhuman”accuracy on the SNEMI3D challenge; a convolutional net bested human accuracy as estimated by disagreement between two human experts (Lee et al. 2017). This research is the basis of the lab’s MICrONS system