Pouya Bashivan, Postdoctoral Associate, Massachusetts Institute of Technology (MIT), Department of Brain and Cognitive Sciences, provides an overview of his work at MIT.
Bashivan discusses the specific area of brain activity that is his current focus. Bashivan states that their lab’s primary area of study is in regard to how the brain sees—the visual cortex, recognition of objects, etc. To put it in perspective, remarkably, Bashivan states that humans can recognize tens of thousands of objects within a few milliseconds. He discusses patches in the brain where neurons respond to objects.
Bashivan, in collaboration with Kohitij Kar, are the lead authors of a new paper that has created quite a stir in the scientific community. The MIT neuroscientists performed the most intensive testing to date of computational models that imitate a brain’s visual cortex. By utilizing their advanced modeling of the brain’s visual neural network, Bashivan and Kar designed a unique way to accurately control individual neurons as well as populations of neurons within the middle of the specific network. Their team’s findings demonstrated that information gained from this computational model would allow them to design images that significantly activated selected brain neurons, specifically of their own choosing. Their groundbreaking research indicates that current versions of these models are similar enough to the brain that they can possibly be utilized to control brain states within animals.
The neuroscientist discusses the next stage and some of the challenges in development for machine learning and artificial intelligence (AI). Further, Bashivan discusses self-driving cars and other innovative technologies, and the complexities involved in regard to vision in their operating systems. He discusses how artificial vision systems can be maximized to operate even better than how human vision operates.
Continuing, Bashivan discusses the importance of focusing on single, larger neural networks as a means to achieve current goals, accomplish behaviors, etc. Additionally, MIT neuroscientist discusses the future of machine vision, going beyond object recognition. He talks about some of the new models that are being trained to play digital games, and this could lead to training in more realistic style games that could approximate real-life situations that humans encounter—vision training as well as brain training. This type of training could provide data on how human brains are impacted by experiences.
Podcast: Play in new window | Download | Embed