‘Organismoid’ device replicates human thought processes
Image credit: Dreamstime
Researchers at Purdue University, West Lafayette, have developed a new technology called ‘organismoids’, devices created from a ceramic material that mirror animal memory by learning to forget unimportant memories while retaining important ones.
“The human brain is capable of continuous lifelong learning,” explained Professor Kaushik Roy of Purdue University. “And it does this partially by forgetting some information that is not critical […] What we are trying to do is mimic that behaviour of the brain to a certain extent, to create computers that not only learn new information but that also learn what to forget.”
Purdue University researchers collaborated with material scientists, electrical engineers, computer scientists and physicists from other US research universities and national laboratories in order to develop the devices, which mimic human thought processes.
The basis of the device is a ceramic quantum material – samarium nickelate – which is used to create organismoids. When exposed to hydrogen, its lattice is doped by hydrogen atoms, and the material undergoes an enormous change in resistance. These reversible, tuneable changes are compared to breathing: it expands with the addition of hydrogen and contracts with its removal.
The organismoid’s changing resistance is similar to a vital feature of animal behaviour known as habituation. According to the researchers, even some organisms without brains possess this.
If an animal comes across certain information regularly – such as the location of food – it is habituated, causing a memory to be retained. Otherwise, the memory begins to decay.
“The behaviour of conductance going up and down in exponential fashion can be used to create a new computing model that will incrementally learn and at the same time forget things in a proper way,” said Professor Roy.
Professor Roy and his colleagues used this behaviour to develod a neural learning model which they named “adaptive synaptic plasticity”. They were able to use the organismoids to implement this model.
“Using this effect, we are able to model something that is a real problem in neuromorphic computing,” said Professor Roy. “For example, if I have learned your facial features I can still go out and learn someone else’s features without really forgetting yours. However, this is difficult for computing models to do. When learning your features, they can forget the features of the original person, a problem called catastrophic forgetting.”
This achievement – detailed in Nature Communications – is significant because it is one of the first examples of using quantum materials directly to solve a major problem in neural learning.
Organismoids could have applications in spintronics, the emerging field concerned with the spin of electronics in solid-state devices, and could result in circuits which mimic biological neurons and synapses in a compact design.
“These devices possess certain characteristics of living beings and enable us to advance new learning algorithms that mimic some aspects of the human brain,” said Professor Roy. “The results have far-reaching implications for the fields of quantum materials as well as brain-inspired computing.”