Researchers Harness the Power of Networked Brains in Monkeys and Rats
Neurobiologists have shown that brain signals from multiple animals can be combined to perform certain tasks better than a single brain.
Brain-machine interfaces could help give disabled people more independence.
New research proves that two heads are indeed better than one, at least at performing certain simple computational tasks.
The work demonstrates for the first time that multiple animal brains can be networked and harnessed to perform a specific behavior, says Miguel Nicolelis, a professor of neurobiology and biomedical engineering at Duke University and an expert in brain-machine interfaces. He says this type of “shared brain-machine interface” could potentially be useful for patients with brain damage, in addition to shedding light on how animal brains work together to perform collective behaviors.
Nicolelis and his colleagues published two separate studies today, one involving rats and the other involving monkeys, that describe experiments on networks of brains and illustrate how such “brainets” could be used to combine electrical outputs from the neurons of multiple animals to perform tasks. The rat brain networks often performed better than a single brain can, they report, and in the monkey experiment the brains of three individuals “collaborated” to complete a virtual reality-based task too complicated for a single one to perform.
To build a brain network, the researchers first implant microwire electrode arrays that can record signals as well as deliver pulses of electrical stimulation to neurons in the same region in multiple rat brains. In the case of the rat experiment, they then physically linked pairs of rat brains via a “brain-to-brain interface” (see “Rats Communicate Through Brain Chips”). Once groups of three or four rats were interconnected, the researchers delivered prescribed electrical pulses to individual rats, portions of the group, or the whole group, and recorded the outputs.
The researchers tested the ability of rat brain networks to perform basic computing tasks. For example, by delivering electrical pulse patterns derived from a digital image, they recorded the electrical outputs and measured how well the network of neurons processed that image. In another test, the researchers delivered information about barometric pressure and temperature and the brain network computed the probability of rain. The brain networks were consistently better than a single brain, especially when the task involved more than one computation step.
In the monkey experiment, the researchers combined two or three brains to perform a virtual motor task in three dimensions. After implanting electrodes, they used rewards to train individual monkeys to move a virtual arm to a target on a screen. An individual monkey brain does not have the capacity to move the arm in three dimensions, says Nicolelis, so each monkey learned to manipulate the arm within a certain “subspace” of the virtual 3-D space. The larger task cannot be completed unless at least two brains work together and achieve a relatively high level of synchronization, he says.
The researchers placed three monkeys in separate rooms with screens, recorded electrical outputs as the animals performed their respective tasks, and then used a computer to combine the outputs. Even though the monkeys didn’t know they were collaborating, says Nicolelis, their brains became synchronized very quickly, and over time they got better and better at moving the arm.
Nicolelis says the phenomenon that led to this synchrony may have important biomedical implications. Shared brain machine interfaces like those demonstrated here will allow “new horizons for clinical applications to open up,” he says. For example, he suggests, perhaps neurologically disabled people could share healthy brain activity from others and collaboratively perform virtual reality-based neurorehabilitation training exercises.
The AI revolution is here. Will you lead or follow?
Join us at EmTech Digital 2019.