It is increasingly common to read or hear about machine learning or machine learning, a branch of artificial intelligence that basically make machines “learn” alone.
Advances in this discipline have been coupled with the use of huge amounts of data to train these engines and achieve surprising results in many cases. Are there limitations to what automatic learning can achieve? Of course, but all these examples show that there are areas in which this discipline could change our world forever.
The fight against suicide
A team from Cincinnati Children’s Hospital is working on a system that after interviewing a number of people trying to figure out if they were people with some risk of suicide.
The analysis not only took into account the answers, but also other sections such as the intonation or the harmony of the answers, something that according to the experts could contribute much information to the system. According to the study, people were identified with suicide risk with 93% accuracy.
How to Speak English (Non-native)
The speakers of the same language tend to be without problems, but there are situations in which those who have learned to natively and those who have done as a secondary language have problems communicating.
Precisely in this area, it is where the MIT created a large base of data in non-native English phrases. The goal was to improve the processing of language machines. Technology language processing (NLP) can work something curious: a machine does not know how to process the particularities of non-native English.
Infallible medical diagnoses (or not)
One area in which it was expected that these systems achieve significantly impact is medicine, and specifically the medical diagnosis in which machine learning theory seemed very capable of giving valid responses.
The self-diagnostic tests conducted by Harvard University showed that automatic systems work with objective data (radiographs, CT scans, clinical analysis) but not with symptoms offered directly by patients.
The problem is that doctors are able to detect and extract those symptoms from patients information (“everyone lies,” said Dr. House) to effectively diagnose what happens to them. A machine will still have time to solve this problem, it seems.