Australian robots invent their own language
A pair of talking robots, initially not possessing any language for communicating with each other, developed it in the course of interaction. The droids gradually agreed and determined that, for example, the word "jaya" means "in the center of the room."
A fun experiment was conducted by Ruth Schulz and her colleagues from the University of Queensland. They investigated the process of language formation from scratch using two machines as an example. The project received the name "Lingodroids" (Lingodroids).
According to Discovery, the cars examined a small area using cameras, laser rangefinders and sonars. In the course of intelligence, robots came up with words to denote certain areas of the terrain.
At first, these words did not exist at all. Each robot compiled them at random, combining syllables in random order.
When the two devices met in one place, the first of them, which had already opened this section, pronounced its name out loud, and the second understood what they wanted to tell him. Further, in a similar way, the vocabulary of constantly communicating machines grew and replenished with terms meaning directions and even distances.
This language was strengthened with the help of games. Let's say one robot pronounced “kuzo,” and then both rushed to where they thought “kuzo” was. If two cars met at the same point or near each other, this success strengthened the connection between the word and the location.
After several hundred games, the robots agreed on directions with an accuracy of 10 degrees and distances within 0.38 meters.
Using the language they invented and personal experience, the robots built each of their own maps of the area, including sections that they could not reach. And these cards turned out to be very similar to each other. Moreover, if some part of their tiny world was behind the door, it also got a name and subsequently, when the door was opened, one robot could ask the second to go there and the partner correctly interpreted what they wanted from him.
In the future, scientists intend to push the “Lingodroids” to complicate the language so that they learn to explain one another - how to get to the indicated point or to characterize the availability of the place.
According to the Australians, this experiment will serve as a platform for the study of cognitive processes related to the representation of knowledge, planning, language development, the binding of symbols to objects and imagination. And such studies will help to improve the means of communication between robots among themselves and with people.
The results of the experiment (PDF document) were presented by its authors in mid-May in Shanghai at a conference on robotics.
Sources: One , Two
Upd. Moved from closed "The future is here" to "Artificial Intelligence"
A fun experiment was conducted by Ruth Schulz and her colleagues from the University of Queensland. They investigated the process of language formation from scratch using two machines as an example. The project received the name "Lingodroids" (Lingodroids).
According to Discovery, the cars examined a small area using cameras, laser rangefinders and sonars. In the course of intelligence, robots came up with words to denote certain areas of the terrain.
At first, these words did not exist at all. Each robot compiled them at random, combining syllables in random order.
When the two devices met in one place, the first of them, which had already opened this section, pronounced its name out loud, and the second understood what they wanted to tell him. Further, in a similar way, the vocabulary of constantly communicating machines grew and replenished with terms meaning directions and even distances.
This language was strengthened with the help of games. Let's say one robot pronounced “kuzo,” and then both rushed to where they thought “kuzo” was. If two cars met at the same point or near each other, this success strengthened the connection between the word and the location.
After several hundred games, the robots agreed on directions with an accuracy of 10 degrees and distances within 0.38 meters.
Using the language they invented and personal experience, the robots built each of their own maps of the area, including sections that they could not reach. And these cards turned out to be very similar to each other. Moreover, if some part of their tiny world was behind the door, it also got a name and subsequently, when the door was opened, one robot could ask the second to go there and the partner correctly interpreted what they wanted from him.
In the future, scientists intend to push the “Lingodroids” to complicate the language so that they learn to explain one another - how to get to the indicated point or to characterize the availability of the place.
According to the Australians, this experiment will serve as a platform for the study of cognitive processes related to the representation of knowledge, planning, language development, the binding of symbols to objects and imagination. And such studies will help to improve the means of communication between robots among themselves and with people.
The results of the experiment (PDF document) were presented by its authors in mid-May in Shanghai at a conference on robotics.
Sources: One , Two
Upd. Moved from closed "The future is here" to "Artificial Intelligence"