A group of scientists have reportedly developed a DNA-based neural network that works exactly like the human brain, only digitally, and can recognize human handwriting.
The experiment, which is currently restricted to the confines of a laboratory test tube, demonstrated its ability to correctly identify handwritten numerals. The researchers who led the project say that’s a significant step toward developing the ability to program artificial intelligence into entirely organic material.
DNA, which is composed of smaller molecules called nucleotides—Adenine, Thymine, Guanine, and Cytosine—is the basic building block of all organic matter. While Adenine will only bond with Thymine, and Guanine only with Cytosine, this creates predictable patterns that can be manipulated in the laboratory.
In 2011 this process was used to develop a more simplistic artificial neural network that could recognize four simple image patterns. The California Institute of Technology team came up with a neural network that is a bit more complex.
The Caltech researchers used their own “designed” DNA sequences that could initiate chemical reactions to indicate it had identified a particular piece of “molecular handwriting.” The neural network could classify up to nine handwritten digits—1 through 9—and was put to the test with 36 different handwritten numerals, all of which it correctly identified.
For instance, when the AI identified a five, it would output green and yellow chemical reactions. For a nine, it would output green and red.
Lead researcher Luly Qian said:
“Though scientists have only just begun to explore creating artificial intelligence in molecular machines, its potential is already undeniable.
“Similar to how electronic computers and smart phones have made humans more capable than a hundred years ago, artificial molecular machines could make all things made of molecules—perhaps including even paint and bandages—more capable and more responsive to the environment in the hundred years to come.”
According to Caltech sources, the team will now develop AI neural networks that can learn, thereby “forming memories” while in their test tube environment. This so-called “smart soup” will then be trained in how to perform specific tasks.