It's not uncommon to hear someone ask their mobile music app to play songs they don't know, but they're confident they'll like the offerings. It's also no surprise that the GPS system provides the perfect route, the result of processing thousands of signals that tell you when other drivers are stopped for an accident or speeding along an alternate route.
The examples are accumulating and Artificial Intelligence (AI) is in many places: in spam filtering, in the automatic translation of content in different languages and in spell checking. However, among the examples with which citizens interact on a daily basis, Uruguayan devices are appearing more and more frequently.
"An important example is the cameras that were installed by the Ministry of the Interior. There is no need for 100 operators viewing the monitors because AI techniques detect in which camera an anomalous event is taking place," explained computer engineer Sebastián García Parra, director and co-founder of Idatha, to 180.
Loan approval also relies on machine tips. The device is trained with historical data of loans granted, with socio-demographic characteristics of customers and the return of payments in each case. "With the past data, an algorithm is built that will predict the probability that a new client will pay the loan requested," García Parra said.
From past data, the machine is able to predict the future. So the more loans the institution makes, the more data it will have and the better decisions its model will make.
Uruguayan examples are growing every day. Virtual assistants in web pages (banking, television for subscribers, pharmacy chains), where the client has the possibility of writing queries in a chat box and begins to interact with an agent available 24 hours a day, are one of the fields in which more growth is registered.
García Parra related the challenges implied by such a system: "Users can express themselves in many ways because written language allows as many combinations as human imagination. Then, we have to train the machine with different ways of expressing the same intention and we achieve it with Automatic Learning. We train it with hundreds or thousands of examples of expressions that communicate the same intention. The more examples we use, the better the customers will understand.
The automatic image recognition transcended by the security measures implemented in football but there are more examples: identify the load level of a truck from the analysis of images, read the license plate of moving cars using surveillance cameras, determine the quality of a cut of meat, detect an anomaly in a magnetic resonance image, are all examples of AI working in Uruguay.
"Our country is increasingly incorporating this technology, both privately and publicly. But it is relevant to understand that these systems work when the tasks are repetitive and when we have good quantity and quality of data to train them. Machines are far from supplanting human beings in creative tasks and being able to think. There is no magic behind this concept. The marketing apparatuses of big companies tend to throw too much fireworks," García Parra concluded.
Source: 180
Connect