When reasoning and planning, humans make use of a wealth of semantic knowledge: place a human in a new house and ask them to make a coffee and they will instinctively search out the kitchen, locate a kettle, etc. We know how to do this as we have acquired a rich set of knowledge throughout our years interacting, learning, and reasoning about the world. We can also explain our thought process to other humans such that they are able to understand and correct our reasoning. Current AI systems severely lack the above: they fail to incorporate semantic knowledge and their world models are often uninterpretable black boxes.
My PhD aims to explore the extent to which natural language can address these shortcomings. Namely: (1) can reasoning about the world in natural language help incorporate rich semantic knowledge into AI systems; (2), can we build a world model with natural language that is interpretable, robust, and verifiable?
I chose to study at the STAI CDT as I believe it offers the perfect environment for a successful, rewarding, and enjoyable PhD, and the theme of Safe and Trusted AI is as important now as it ever was, as AI systems become more ingrained in everyday life.
–
Masters Qualification: MSc Computing Science Imperial College London
Undergraduate Qualification: BSc Economics University of Surrey
Work Experience:
- Data Scientist, Mantle Labs
- Fast Stream Economist, UK Government
LinkedIn: https://www.linkedin.com/in/matt-barker-897007151/
