Artificial intelligence (AI), and it is not necessary to give examples at this point, is already capable of seeing, hearing, speaking, … But now, in addition, and thanks to IBM it is also capable of tasting thanks to Hypertaste, an electronic language which is capable of detecting, testing and analyzing liquids quickly. Food, health or environmental management are some of the fields in which this solution can have multiple uses.
For years, technological innovation has focused mainly on the senses of sight and hearing, allowing them to be significantly improved with devices nowadays available to anyone; glasses, microscopes, speakers, or devices that correct hearing loss. In the field of artificial intelligence, it has also been the senses of sight and hearing that, so far, have experienced the greatest advances. There are, for example, computer vision systems that are capable of learning to recognize objects or faces after a relatively short training period.
Likewise, there are also sound and speech recognition systems that we use daily: from virtual assistants on our mobile phones, to applications capable of identifying the song we are listening to at a specific moment. However, innovations related to the chemical senses such as taste and smell have not experienced the same level of technological progress. A fact that, to a large extent, can be caused by the enormous complexity of these senses and the relatively scarce scientific knowledge of the physical-chemical processes that govern them.
Taste, a complex and little explored sense in artificial intelligence
Responding to this need for innovation, IBM Research has developed Hypertaste, an ‘electronic language’ assisted by artificial intelligence that, inspired by the functioning of the human sense of taste, will allow the detection and analysis of different types of liquid quickly and without the need to go to a laboratory . Hypertaste is a small circular device that is partially introduced into the liquids to be analyzed. To perform the analysis and detection of a fluid, this ‘electronic tongue’ uses IBM’s AI and a set of multisensitive electrochemical sensors, each consisting of pairs of electrodes.
Complex liquids contain many different molecules and it is the combination of all of them that distinguishes them from each other. Therefore, in order to identify them, it is necessary to analyze their molecules as a whole. Just as the sense of taste or smell does not have a receptor for each molecule in a food, but rather reacts from a specific combination, Hypertaste sensors are capable of responding simultaneously to different chemical compounds. In this way, through the combination of these different sensors, a “holistic” analysis of the set of components of a liquid can be carried out and its “fingerprint” can be found.
All the data collected is transferred through a mobile application to the cloud, where a machine learning algorithm compares this fingerprint with a database with information on known liquids. The algorithm determines which liquids in the database are chemically similar to the liquid under investigation. The result obtained is sent directly to a smart device, such as a mobile phone.
A very fast device
One of the advantages of Hypertaste is that it is extremely fast. This entire identification process takes less than a minute, from the moment the device is in contact with the liquid until it delivers a result. This is record time, especially considering that conventional laboratory systems often require hours. In addition, due to its small size, it is a portable and easy-to-use tool.
A key aspect in this whole process is the “training” received by the machine learning algorithm that Hypertaste uses. As with a sommelier , who with time and practice learns the complexity of tasting wine, Hypertaste training is essential so that you can identify the different liquids. In the case of Hypertaste, this “training” consists of measuring many times the set of voltage signals produced by the device’s sensors. Each sensor (that is, each pair of electrodes) reacts to the dissolved ions in the studied liquid with a voltage signal. The fingerprint of a liquid is nothing more than the result of the combination of all these voltage signals that the liquid in question generates in the Hypertaste sensors. In short, the task of training Hypertaste’s AI algorithm focuses on “teaching” it to recognize the characteristic pattern of voltage signals of a specific liquid through repeated exposure (across multiple measurements) to that pattern.
Applications
Hypertaste can be a very useful solution for both industries and scientists, as it can help identify liquids in a fast, portable and reliable way, without the need to have access to high-end laboratories. The pharmaceutical and healthcare sector are just one example of the industries that could benefit from the capabilities of this solution, as its capabilities span many more areas. Thus, for example, it can be very useful for an organization that needs to check the water quality of a river or lake. Or for a manufacturer who wants to verify the origin and status of the raw materials that he will use to make his products. Or for a producer trying to identify counterfeit drinks, such as whiskey or wine.
In the future, Hypertaste could even detect the fingerprint of other even more complex liquids. In the long term, it could, for example, take urine samples from a person and help to obtain an assessment of the metabolic fingerprint, which can be understood as the sum of all the small molecules present in a living organism. As this chemical information is constantly changing (depending on factors such as lifestyle or nutrition) this metabolic fingerprint could help to have a “snapshot” of a person’s health at any given time.
The potential applications and fields of work for Hypertaste are many, and teams at the IBM Research laboratory in Zurich are currently working to study how to get the most out of this innovative solution, powered by artificial intelligence.
