Some of the most impressive things that artificial intelligence is able to do have to do with your ability to analyze data, Communicate fluently or identify patterns. Although there are still many techniques to perfect, see, hear and speak, it is no longer a challenge for this technology. However, the sense of smell is resisting. Among humans, it is one of the most underrated senses, but in animals such as dogs it is especially useful in aspects such as the detection of diseases. These animals have become a reference for different international projects that are developing the olfactory capacity of artificial intelligence.
There are dogs that can detect diabetic episodes hours in advance, the early onset of melanoma and even breast, lung, colorectal, ovarian and prostate cancer, which leaves a trace in the man's urine. In Wired you can read the story of the Dogs of Medical Detection, a group of dogs from the United Kingdom that specializes in detecting this type of ailments through their nose. Taking into account all the usefulness of this sense, how is it possible that we still can not use machines that smell, as we use machines that listen and see?
One of the main problems is that you do not know too much about smell: for the moment it is the most mysterious sense. What we do know is that we have chemical sensors that are excited by air particles and that produce signals that are interpreted by the brain. But there is no classification of odors and it is very difficult to describe an odor or name it without associating it with a specific object, saying, for example, that something "smells like freshly cut grass". This difficulty in cataloging smells makes it more complicated to translate them into the language of machines.
Therefore, most objects that can smell today are, in reality, detectors. This is what happens with smoke detectors: they identify the chemical they have pre-established but they are not able to interpret the information they collect. For example, they do not analyze data such as what type of smoke it is, if it comes from something that is on fire or if it is a cigarette, how long it has been burning or how much water would be necessary. These aspects would make the detector an intelligent device.
One step beyond are the so-called electronic noses. NASA developed its own already in 2004 to apply it in space exploration. It was a device that was programmed to learn to recognize "almost any compound or combination of compounds. You can even train to distinguish between Pepsi and Coca-Cola, "then assured NASA. In this case, it was being designed to detect when ammonia levels reached dangerous limits for astronauts or to predict a fire. The US space agency says that "represents an interesting tool for various applications, such as control of food quality, the identification of harmful gases and biomedicine."
Currently, efforts are focusing precisely on the health sector and the possibility of using an intelligent device that is capable of smelling diseases as dogs do. To this end, Andreas Mershin, physicist and director of MIT's Label Free Research Group, and his colleague and mentor, Shuguang Zhang, are training an artificial intelligence system with a database obtained from dogs that are experts in detecting diseases. through smell
The goal is to create a device called Nano-Nose. That information will help them select which receivers they need to place on the device. From there, the machine can continue learning about your answers. The idea that reaches a point where no one can differentiate between the reactions of the device and those of a dog. "If all goes well, the Nano-Nose will become more than just a detection device; will be a diagnostic tool, "the researchers explain. Wired.
The culmination, according to Mershin, would be to see the Nano-Nose incorporated in a smartphone. If this device could be incorporated into mobile phones, which are in constant contact with the body of users, data on the health of the owner could be collected. "The artificial nose could alert you to being checked on the mole of your thigh or warn you that your blood sugar level is going down dangerously." The debate that arises on this point is already omnipresent: how would this information be managed? Probably your insurance would pay to know exactly what your state of health is.
(tagsToTranslate) objectify (t) machine (t) be (t) able (t) sniff (t) precision (t) dog (t) can (t) power (t) detect (t) disease (t) smell ( t) point (t) reference (t) different (t) project (t) international (t) intelligence (t) artificial