Artificial Intelligence Tool Aims to Identify, Predict Smells
2022-09-15
LRC
TXT
大字
小字
滚动
全页
1American researchers have created a technology tool that aims to identify and re-create different smells.
2The system is powered by artificial intelligence (AI).
3The tool was developed and tested by researchers at Google, universities and the Monell Chemical Senses Center.
4The team recently released results of their method and described the process in an article in the publication Proceedings of the National Academy of Sciences.
5The researchers said smells, also called odors, are produced by molecules released into the air.
6Those molecules then enter our noses and are processed there by "sensory receptors."
7Those receptors send signals to the brain to help us recognize smells.
8The team said smells are identified in a way similar to how we identify and measure colors - using sensory maps.
9For example, maps - like the traditional color wheel - can show many different colors.
10Such maps present color combinations and show how the colors blend with each other.
11But the researchers said that historically, "useful maps for smell have been missing."
12This is because the molecules related to smell are much more complex.
13They depend on more than photon particles used to see colors.
14In addition, the researchers said that while the human eye has just three sensory receptors for color, the human nose has more than 300 for smell.
15The scientists built on past research to produce a neural network model that aims to create a "map" to identify molecules related to smells.
16A neural network is a computer processing system built to act like the human brain.
17The team from Google AI calls the newly developed tool a "principal odor map."
18The main purpose of the map is to "predict the odor properties of molecules," the researchers said.
19The neural network models were also trained to identify which molecular characteristics were needed to make correct smell predictions.
20The researchers reported that the AI-powered map permits them to "predict and discover new odors and the molecules that produce them."
21The team said their research suggests there are probably billions of molecules that are likely odorous, but have not yet been smelled.
22Since the mapping tool was shown to recognize an odorous molecule from its structure, "this guides us to discover new classes of odorants..." the researchers said.
23For the experiments, the researchers said they collected the largest ever set of data on odor descriptions for different molecules.
24For comparison purposes, they used human subjects to identify the smells of 400 molecules using 55 different descriptions.
25The team said that when compared to the human odor identifications, the neural network model repeatedly performed better than the human subjects.
26The team said such a tool can be used across many different fields, including the health care, food and fragrance industries.
27The researchers reported that the tool can be used to help fight one of the world's biggest health problems - diseases spread through insects.
28They said the system can measure how effective a molecule is at keeping mosquitos away from humans.
29The team said it discovered a series of new molecules that can repel mosquitos with at least the same effectiveness as DEET, a chemical commonly used in insect repellants.
30This discovery, the researchers said, can lead to the development of less costly, longer lasting, and safer repellants than those containing DEET.
31Such repellants could be used "to reduce the worldwide incidence of diseases like malaria, potentially saving countless lives," the team added.
32I'm Bryan Lynn.
1American researchers have created a technology tool that aims to identify and re-create different smells. 2The system is powered by artificial intelligence (AI). The tool was developed and tested by researchers at Google, universities and the Monell Chemical Senses Center. The team recently released results of their method and described the process in an article in the publication Proceedings of the National Academy of Sciences. 3The researchers said smells, also called odors, are produced by molecules released into the air. Those molecules then enter our noses and are processed there by "sensory receptors." Those receptors send signals to the brain to help us recognize smells. 4The team said smells are identified in a way similar to how we identify and measure colors - using sensory maps. For example, maps - like the traditional color wheel - can show many different colors. Such maps present color combinations and show how the colors blend with each other. 5But the researchers said that historically, "useful maps for smell have been missing." This is because the molecules related to smell are much more complex. They depend on more than photon particles used to see colors. 6In addition, the researchers said that while the human eye has just three sensory receptors for color, the human nose has more than 300 for smell. 7The scientists built on past research to produce a neural network model that aims to create a "map" to identify molecules related to smells. A neural network is a computer processing system built to act like the human brain. 8The team from Google AI calls the newly developed tool a "principal odor map." The main purpose of the map is to "predict the odor properties of molecules," the researchers said. 9The neural network models were also trained to identify which molecular characteristics were needed to make correct smell predictions. The researchers reported that the AI-powered map permits them to "predict and discover new odors and the molecules that produce them." 10The team said their research suggests there are probably billions of molecules that are likely odorous, but have not yet been smelled. Since the mapping tool was shown to recognize an odorous molecule from its structure, "this guides us to discover new classes of odorants..." the researchers said. 11For the experiments, the researchers said they collected the largest ever set of data on odor descriptions for different molecules. 12For comparison purposes, they used human subjects to identify the smells of 400 molecules using 55 different descriptions. The team said that when compared to the human odor identifications, the neural network model repeatedly performed better than the human subjects. 13The team said such a tool can be used across many different fields, including the health care, food and fragrance industries. 14The researchers reported that the tool can be used to help fight one of the world's biggest health problems - diseases spread through insects. They said the system can measure how effective a molecule is at keeping mosquitos away from humans. 15The team said it discovered a series of new molecules that can repel mosquitos with at least the same effectiveness as DEET, a chemical commonly used in insect repellants. 16This discovery, the researchers said, can lead to the development of less costly, longer lasting, and safer repellants than those containing DEET. Such repellants could be used "to reduce the worldwide incidence of diseases like malaria, potentially saving countless lives," the team added. 17I'm Bryan Lynn. 18Bryan Lynn wrote this story for VOA Learning English, based on reports from Google AI and PNAS. 19_________________________________________________________ 20Words in This Story 21artificial intelligence - n. the development of computer systems with the ability to perform work that normally requires human intelligence 22blend - v. to mix two or more things together 23photon - n. a single unit of light 24characteristic - n. a typical or noticeable quality that makes one person or thing different from others 25fragrance - n. a sweet of pleasant smell 26repel - v. to force away something unwanted 27incidence -n. the number of times something happens 28What do you think of this story? We want to hear from you. We have a new comment system. Here is how it works: 29Each time you return to comment on the Learning English site, you can use your account and see your comments and replies to them. Our comment policy is here.