An artificial neural network made from DNA that is able to solve a machine learning problems especially in identifying handwritten numbers. It is a huge step in for proving the capacity to program artificial intelligence into biomolecule circuits. This network has developed by researchers based at Caltech in the laboratory of Lulu Qian, assistant professor of bioengineering.
Qian said, scientists have begun to test artificial intelligence in molecular machines, and it is proving its capabilities. Likewise, the computer and smartphones have improved the capabilities of human from past few years. Molecular machines could develop from molecules, and developments such as paints and bandages can improve response to the environment in years to come.
These artificial neural network is inspired from human brain and constructed from mathematical models. However, they are simpler than biological parts, artificial neural functions same as neurons and improves behavior. The laboratory is focused to program the behaviors with made from DNA. Human has 80 billion neurons in the brain, which are capable to make a sophisticated decision. Smaller animals can make simpler decisions with just hundreds of neurons. To achieve proper functioning the machines are designed in the simpler way, added Qian.
For initial testing, the task of recognizing handwriting was selected as it changes per person. Recognizing handwriting is a difficult task to do, sometimes human fails to recognize. These AI machines are taught to recognize numbers and handwriting. As the molecular number is made up of 20 DNA chosen from 100 molecules, each of them is assigned for 10 by 10 pattern. These are mixed together and passed the test.
Few biomolecules such as blood glucose and cholesterol via medical diagnostics tests. Using biomolecular circuits that might be available in human, these responses are conducted directly to the molecular environment.