Synthetic Brain to Aid Research
In what is certainly initially seems like a novel twist on logic, researchers from UC Irvine are utilising a robot outfitted with an artificial mind, to help them understand how the human brain works.
Of course the problem with using the human brain to understand how the human brain works, is that the brain is so complicated, it is hard to know where to start, and just as importantly, where to stop. Therefore, UCI cognitive scientist Jeffrey Krichmar and former Hollywood animatronics engineer Brian Cox designed a robot, called CARL, to think and act at least a little bit like a human does.
'Little bit' is right. CARL is running a trimmed down version of a human brain, with only about two million connections - about the size of a mouse brain. Half programmed, and half cognitive AI, the simplified brain has been designed to mimic as accurately as possible, certain sections of a huamn brain. The hope is that this simplified structure will provide insight into the specific brain areas responsible for decision-making and attention, advancing robotic design as well as knowledge of human behaviour.
"Little is known about the areas of the brain involved in making decisions when faced with uncertainty," said Jeffrey Krichmar, a UCI cognitive scientist and one of the study's lead researchers.
Krichmar specializes in neurorobotics, or programming robots with real-life organic and neural data to simulate thinking, moving beings. His work with CARL - a robot with a biologically plausible nervous system controlled by a realistic model of the human brain - has led to several advances in the field, the most recent of which are featured in the September issue of IEEE Robotics & Automation Magazine.
Collaborating with Krichmar are UCSD researchers Andrea Chiba, Douglas Nitz and Angela Yu, who will develop the data for the study by testing the decision-making abilities of rodents during a task in which the locations of stimuli that predict food rewards change abruptly, requiring the rats to adapt to the new environment in order to receive food rewards.
Brain recordings taken from the rodents during the task will be digitally analysed and programmed into CARL's software-controlled "brain," enabling the robot to replicate the same behaviour.
"As the robot navigates the same challenging situations the rats faced, though, we'll be able to actually see the areas of the simulated human brain being utilised to make decisions and the physical changes taking place."
It's not going to be a complete model of course, but even if the experiment does not succeed, because the structure mirrors a simplified version of the human brain structures so closely, we will at least learn where not to look. If it succeeds, we will know where on the brain to focus our efforts. Either way, promises a boon for brain, and thus, AI work.