It seems lightyears since Stanley Kubrick’s epic "2001: A Space Odyssey" (1968) planted the idea in our minds that humans and machines would one day converse. Even if HAL 9000 was hardly the best advertisement, the future of talking AI-equipped counterparts seemed set in the stars. thingsTHINKING, a spin-off from the Karlsruhe Institute for Technology (KIT), is appearing at the Baden-Württemberg International pavilion in Hall 14 during CEBIT 2018 to unveil a potential milestone in this direction. The artificial intelligence onboard its NLP semantic processing platform understands the meaning of language, which is how humans - and potentially now machines - are able to process unstructured information.
To reach this stage of development, the founders of thingsTHINKING, Dr. Mathias Landhäusser and Dr. Sven J. Körner, first needed to realize that machine learning, statistics and other existing approaches to human-machine interaction were never going to do the job. At least not as long as machines process communication in purely mathematical and statistical terms - without actually understanding the meaning of human language. The futility of this entirely logic-based approach is revealed by Ludwig Wittgenstein’s simple yet ingenious explanation that "the meaning of a word is its use in the language". Language is heavily influenced by the communicator's life experience - and a gulf still remains between humans and machines, at least for now. In contrast to existing methods of NLP/NLU (natural language processing/understanding), the semantic processing platform created by thingsTHINKING now seeks to understand, process and use the semantics of human speech. This places it way ahead of the field for a whole range of potential applications. The thingsTHINKING software on show at CEBIT is designed to spot humans’ linguistic anomalies and compensate for them. This is likely to prove immensely useful in Industry 4.0 and for software manufactures or consultants, paving the way for improved customer service using virtual assistants, or legal-tech aids for interpreting contracts, for example.