Robot doctors and oenologists
AI applications that can even predict the quality of a wine being presented at Athens conference
Is there a way to predict the quality of the wine from this year’s grape harvest? Can we know what a court will decide months before its final ruling? Can we help prevent cardiovascular disease in the future? Or, even, can we predict the weather using historical meteorological data? Artificial intelligence is here to provide a positive answer to all these diverse questions.
How AI can be applied and what changes it may bring to our daily lives and to every field of the sciences is set to be discussed by Dimitris Bertsimas, a professor at the Massachusetts Institute of Technology (MIT) at an event organized by the Hellenic Institute of Advanced Studies (HIAS). Taking place in Athens on July 3-7, the goal is to build bridges between scientists and innovators in the diaspora and in Greece. Bertsimas, however, gave us a small taste of what to expect.
The quality of a wine, for example, is something quite distinct and depends on a confluence of factors: flavor, aroma, color etc. At the same time, it also relies on the specific natural and climatic conditions of the location where the grapes are grown. “We developed a system of machine learning that makes it easier to choose wine. We combined the prices of wine from many different years on the reasoning that quality determines cost to a significant degree. We also accounted for a series of other natural factors, such as temperature, precipitation and other indexes that have an effect,” Bertsimas tells Kathimerini. This allowed the team to record which climatic combinations in a particular area led to a higher-quality wine. “In fact, Wine Spectator, which is a specialized American magazine on wine and wine culture, uses our app,” adds the MIT professor.
Predicting which way a court will rule seems even more complicated. “We took the decisions of the United States Supreme Court going back several years, and after processing them put them through machine learning to predict decisions that were coming up. We asked a team of experts the same question and the results were impressive: AI had a better prediction rate of around 15%. It was not perfect – its success rate was around 75% – but it did better than a team of scientists,” says Bertsimas.
‘It’s important for people to see that things which are very qualitative, such as the bouquet of a wine, can be quantified’
Health is another crucial area that stands to be deeply affected by artificial intelligence. “We all know that the likelihood of someone developing heart disease relies on a series of factors, such as cholesterol levels, blood sugar etc. Based on these measurements, we can calculate the risk of a heart attack or other heart disease in any individual and see what they need to do,” Bertsimas explains, stressing the importance of proper intervention. “In this way, we can move to personalized medicine, to specialized individual treatments and, therefore, to more effective treatments,” he says.
How we predict the weather right now, meanwhile, relies on all sorts of very difficult and diverse factors and calculations, but Bertsimas says that even better answers may lie in using data from the past. “Indeed, if we combine the two methods, then the level of predictions will become much more reliable,” he says, commenting on another area where AI will play a crucial role. “This would allow us to make decisions on when to evacuate an area because of a hurricane in the US or a wildfire in Greece,” he says.
By using evidence and big databases from the past, scientists working in AI and machine learning are trying to predict future developments and adopt the necessary measures to either prevent disaster or help a good outcome.
“We present AI applications in unusual areas, with examples from the real world, in order to demonstrate that it is suitable everywhere. “It’s important for people to see that things which are very qualitative, such as the bouquet of a wine, can be quantified,” underlines Bertsimas.
The classes being carried out as parts of the HIAS summer events are addressed to students from all sorts of disciplines and so far 1,700 people have registered to attend, 400 of them in person. “We want to mobilize the youth, to get them interested in how we can use algorithms,” says the academic.
Who does the deciding? “The human decides, not the algorithm. The doctor decides on a treatment, not the machine. Proper cooperation is what we’re after. Machines are better at some things, like calculations, but humans can pose the right questions,” he answers. “There are risks, of course, which is why it’s important to educate young people.”
The distinguished Greek academic sees AI bringing radical changes to the sciences, but also to education. “The university of the future could have horizontal training in AI across all the disciplines, even in the same amphitheater,” he says.
