Towards artificial intelligence with common sense

Mae'r digwyddiad hwn wedi dod i ben

Cardiff University's College of Physical Sciences and Engineering invites you to the next event in our inaugural Lecture Series, delivered by Professor Steven Schockaert from the School of Computer Science and Informatics.

Towards artificial intelligence with common sense

In recent years, remarkable progress has been made in the field of natural language processing, and artificial intelligence (AI) more generally. This progress is largely due to the now widespread use of so-called deep learning methods, a subset of machine learning which is loosely inspired by the way in which information is processed in the human brain.

Despite achieving impressive results on a variety of tasks, these deep learning methods mostly rely on rather superficial forms of pattern matching, and intuitively lack the ability to actually “understand” language. As a result, current state-of-the-art systems have proven to be surprisingly brittle, and even replacing one word in a document by a synonym may lead such systems to drastically alter their predictions.

This is a fundamental problem which can only be solved by giving AI systems access to more knowledge about the world, encoding and learning such knowledge has been the holy grail of AI research since the 1960s. Initial attempts were mostly based on learning knowledge in the form of rules, but this was largely abandoned in the 1990s because of the difficulties in acquiring such rules and the fact that they cannot cope with all the nuances of meaning that are often needed.

Most current approaches are based on geometric encodings which are easy to learn, but are only able to capture knowledge in an approximate and shallow way. These limitations are giving rise to an emerging hybrid approach which aims to combine the best of both worlds, and ultimately offer a solution to this decades-old problem.

**The Lecture takes place from 6-7pm in Cardiff University"s Hadyn Ellis Building, Maindy Road however please join us for refreshments from 5pm in the foyer of the Hadyn Ellis Building, and again after the Lecture.