January 29, 2013
Can Robots Invent Their Own Language?
|In an enlightening lecture last year, artificial intelligence researcher Luc Steels discusses his work with language-learning robots and asks: can machines be creative enough to invent their own language?|
The robots are used in the experiments of Luc Steels, in which the complete chain of intelligence from physical embodiment to language is modeled.
The robots in Steel's tests use their artificial intelligence to find their own words in much the same way our ancient ancestors must have, by naming the actions they perform as well as the things around them. The “words” they invent begin as random sounds given to a specific action, object, or event.
In an enlightening lecture last year, Steels discusses his work and asks: can machines be creative enough to invent their own language?
In the lecture below, Steels talks about some of his recent breakthrough experiments, which have seen robots programmed to play language games and come up with novel concepts, words and meanings.
He discusses how this triggers a process of cultural evolution that leads to more complex forms of language and deliberate on what this tells us about the nature of our own intelligence and the future of artificial intelligence.
Steels' focus in particular is on how categories (like colours) can be grounded in perceptual experience and develop under the strong influence of language, and how grammars and the semantic domains expressed by grammars may emerge in a population of agents. Applications are far reaching, ranging from adaptive communication systems for humanoid robots to evolving ontologies and communication protocols with emergent semantics.
Steels is ICREA Research Professor at the Institut de Biologia Evolutiva (CSIC-UPF) in Barcelona and Director of the Sony Computer Science Laboratory in Paris.