Published in The nature of human behavior The article describes how artificial intelligence can pick up the meanings of even very complex words. Scientists also found a way to access this information.
The team dealing with this case included representatives from the University of California, the Massachusetts Institute of Technology and the National Institutes of Health. The goal of their research was to see if AI-powered computers would be able to simulate thought processes and understand words in a similar way to humans.
The system under study has been used in recent years to analyze the meaning of words. It is able to capture semantic differences by analyzing billions of words downloaded from the Internet into prime factors. An example would be a case where words like “table” and “chair” appear side by side. In such circumstances, AI realizes that their meanings are related to each other. On the other hand, when two words are rarely used together, it becomes clear that their meanings are different.
Computers can classify concepts in a similar way to humans
And while this makes sense, if people only relied on this approach, we might have a lot of trouble understanding and distinguishing words. As a result, the new study’s authors wanted to see if the system they were testing would be able to detect the less obvious differences. If this is the case, it would suggest that the logic of computers and humans is much closer than we might suppose.
In total, the authors looked at 52 word groups to see if the system could learn to sort meanings. It was, for example, about the possibility of classifying animals by their size or the danger they posed to humans. There were many groups and they were certainly different from each other – others concerned, among other things, with sports, legends and competitions. Most importantly, each category has multiple contexts. The comparison group consisted of people who would rate the chosen words in a similar way.
What are the conclusions reached by the scientists? First of all, it turns out that AI is the brain much like humans. The computer categorized the words very efficiently, noting the fact that the names “Betty” and “George” are very old and refer to two genders. On the other hand, weightlifting and fencing are found to be similar in that they are practiced indoors, but differ in the level of intelligence needed to practice them.
“Infuriatingly humble musicaholic. Problem solver. Reader. Hardcore writer. Alcohol evangelist.”