If someone blindfolded our eyes and handed a bouquet of roses, although we could smell the tempting fragrance, we might not be able to tell what it was. So, can artificial intelligence recognize that it is a rose based on the fragrance of the flower? What does a keen sense of smell mean for the future of artificial intelligence?
Artificial intelligence is more sensitive than a human nose and
US researchers at the University of California, Riverside, using machine learning to decipher the smell of chemicals. This is a scientific research funded by the National Science Foundation and has potential applications in the food and flavor industries.
Ananda Sankar Rey, a scientist involved in this project, said: “In everyday products, some scented substances may be toxic or irritating. We need to find safer, more natural or mild substances to replace them. , We can use artificial intelligence to predict the odor of natural or artificial chemical substances.”
Some people may suggest that we can hire people with olfactory expertise to do this work. However, compared with artificial intelligence, human work efficiency and accuracy are too low. Artificial intelligence can not only screen hundreds of chemical substances per minute, but also provide relatively accurate odor parameters, which cannot be compared with people with the most sensitive noses. Moreover, allowing humans to identify a variety of odors with uncertain safety may also be harmful to health.
How to make artificial smell
that humans can smell the odor of a substance, because the human nose has five million olfactory cells, about 400 odor receptors. When the volatile molecules of a substance stimulate the odor receptors, these receptors are activated, and the human body can feel the smell of the substance. To unravel the mystery of smell, it is necessary to figure out the receptors corresponding to different smells. At present, physiologists have fully understood the principles of human smell.
The University of California research team adopted a bionics research method, based on the principle of human olfaction, using the research results of informatics and machine learning to establish an artificial intelligence olfactory model. The researchers first developed an intelligent program that allows machines to learn how human odor receptors operate. Then, they arranged about 1 million chemical substances for the artificial intelligence sense of smell for odor testing, and constantly adjusted the parameters until the artificial intelligence sense of smell could be accurately measured. In recent tests, the artificial intelligence sense of smell can distinguish the scent of roses, and can also judge the varieties of roses based on the subtle differences in scents. When horticultural scientists develop new flower varieties in the future, fragrance indicators can be provided by artificial intelligence testing.
Forecast smell of new substances
artificial intelligence in addition to smell the smell of existing substances screening, but more important is to predict the smell of new substances. In other words, it is impossible for us to know exactly the smell of a substance before it is synthesized or discovered. However, artificial intelligence can predict the smell of new substances based on various parameters already in the olfactory database.
For example, a company hopes to develop a perfume with a special scent. After a laboratory simulation, researchers propose 100 new substances as candidate fragrance ingredients. The artificial intelligence sense of smell will screen these new substances that have not yet been synthesized, and finally find new substances suitable for making perfume.
The artificial intelligence prediction of the smell of chemical substances has created a new scientific method. This is of great help to chemists and light industry new product developers. After they develop new chemical substances or products, they can use artificial intelligence to determine whether these substances can be used in food, spices or other related industries.
In addition, the artificial intelligence sense of smell can also test the odors produced by a variety of chemical substances combined according to a certain formula. Therefore, the artificial intelligence sense of smell provides scientists with a wide range of scent “palettes” from which they can formulate scents for various functions. For example, scientists have used this method to develop a mosquito repellent that will keep mosquitoes away, but it is tasteless and harmless to humans.
Let the robot distinguish the smell
To make robots more like humans, it is necessary for robots to obtain all the senses of humans, among which sight, hearing, touch, taste and smell are the most basic five senses. At present, there is basically no problem with the robot’s hearing and vision, and the requirements for finer touch, taste and smell are still far away. The artificial intelligence olfactory developed by scientists can be installed as a smart program in a smart odor detector or in a robot. Therefore, future robots can distinguish various smells.
If the robot has a sense of smell, then it can be used in many fields. In daily life, it can better help people cook fragrant food, and can also help people choose more suitable daily necessities, or keep people away from dangerous areas. In the field of anti-terrorism security inspection, it can use special smell molecules emitted by explosives to help special police find hidden explosives. In the field of criminal investigation, robots may be more professional than police dogs in smelling and tracking fugitives. In the field of disaster prevention and mitigation, it can track the source of harmful gases, such as helping people find the location of gas pipeline leaks.
”Hi! Bring me a bottle of rose-flavored perfume.” One day in the future, when shopping in a mall, a robot next to you may suddenly say something like this. You will probably not be surprised by it, but will smile because of that. Robots with a sense of smell have become more and more popular.
