Does AI really understand the language or only uses it?
Title: Meaning mining in the context – does the AI really understand the language or only uses it?
Natural language processing as a field is offering more and more efficient methods for several complex text-related tasks. Most of them are based on vector representation of words, phrases or sentences. The outcomes are often amazing – and smoothly passing the Touring Test. Although does successfully operating on language mean that the AI does understand the language? That it knows the meaning of the words or sentences it plays with? And how can the notions of meaning and understanding be defined here? With the help of semiotics and philosophy of language I will discuss the concept of word embeddings algorithms, as well as its consequences for the use of linguistic data.
Inez Okulska is a researcher at NASK National Research Institute. Working on intersections of AI and humanities, especially focused on Natural Language Processing (NLP). Having her PhD in literary translation, she is about to graduate from Master of Science in Robotics and Automation. Active literary critic and translator, among others for Scientific American Journal. From May till November 2014 she was conducting her research as a visiting researcher at Harvard University, Cambridge, MA.