Often, when there is a need to study how a particular part of the body works, researchers can create a lab version of an organ so that a process can be observed and possibly manipulated. Projects that include circulatory systems, brain tissue and even tiny bits of heart muscle have all yielded good results.
The latest in lab created material is what scientists at Purdue have dubbed "organismoids, " and they are being used to further the study of artificial intelligence (AI). For a computer to mimic how the human brain thinks, there must be a way to translate the brain processes that surround learning to algorithms.
Kaushik Roy, Purdue University's Edward G. Tiedemann Jr. Distinguished Professor of Electrical and Computer Engineering recently published a study on how to use organismoids for part of this process. He explained, "The human brain is capable of continuous lifelong learning. And it does this partially by forgetting some information that is not critical. I learn slowly, but I keep forgetting other things along the way, so there is a graceful degradation in my accuracy of detecting things that are old. What we are trying to do is mimic that behavior of the brain to a certain extent, to create computers that not only learn new information but that also learn what to forget."
Re-creating the process of learning new information and accurately deciding what to forget, in the realm of AI, is what Roy and colleagues from Rutgers University, MIT, Argonne National Laboratory and Brookhaven National Laboratory have been working on, using a material called "samarium nickelate to make the tiny organismoids.
While not actual living tissue, the material does react when manipulated. Exposing it to hydrogen gas causes the lattice structure of it to expand and then contract when the hydrogen atoms are taken away. It's almost as if the material "breathes." What's happening is that it's undergoing a change in electrical resistance when hydrogen is added, and the effect reverses when the hydrogen goes away. Finding a material that has "quantum" properties like this is unusual but could provide a way forward to understanding the brain process of habituation. Habituation is what happens when we see a behavior repeated often enough that we learn it. If the frequency of how often it occurs decreases, the brain forgets it, making way for learning newer information. That brain process is quite similar to what Roy and the other scientists involved in the work have recreated with these organismoid chips. Being able to orchestrate this process in the lab is key to eventually being able to create algorithms for it for use by AI computers.
The paper details what is essentially a lab created "neural learning model" which the authors call synaptic plasticity. Shriram Ramanathan, a Purdue professor of materials engineering and co-author of the work, explained, "This could be really important because it's one of the first examples of using quantum materials directly for solving a major problem in neural learning."
The goal for the research team was to be able to use the manipulation of the samarium nickelate to model habituation in neuromorphic computing which is involved in applications like facial recognition, decision making, and reasoning. Getting neuromorphic computing close to how the human brain works is the gateway to making AI work properly. Check out the video below to learn more about this incredibly complex process that could change how computers model human thought.