My lab studies visual recognition in the primate brain, as implemented by the electrophysiological activity of neurons in different areas of cortex. Our ultimate goal is to develop a theory of visual recognition that accounts for the activity of neurons in the context of natural environments, not just under laboratory conditions. This is important because of the rise of automated vision technologies, such as convolutional neural networks and transformers, which are proving increasingly useful in industries like manufacturing, transportation and medicine. These artificial models of vision are still vastly limited compared to the brain, and I believe that defining the strategies used by the visual brain will provide opportunities to improve these models. To define how neurons function in naturalistic settings, we use tools from machine learning, including new classes of models called generative adversarial networks and transformers. These neural networks can synthesize images de novo, using compact number lists (vectors) as inputs. We led the first effort that linked these image generators to the activity of living neurons in the macaque brain, allowing neurons to guide the synthesis of their preferred visual stimuli, based on their internal code. This is a powerful new approach that has revealed many exciting forms of visual stimuli that had never been used in the literature before, and will likely fill gaps in our view of neural coding. We are now working to expand this approach to associative cortical regions like prefrontal cortex and subcortical regions like the amygdala.