16 May 2019
Talk 1: Applications for TensorflowJS to improve the customer experience
- Mika Rehman
Synopsis: TensorflowJS has real world applications for websites that could benefit from machine learning. While computer vision and CNN's get a lot most of the glory, the less sexy elements such as word2vec and linear regression can be used to make a real difference to customers experience today. Using machine learning to pre-empt customer contact, using machine learning to understand what a user was looking for when they hit a 404 page and not miss out on sales, using machine learning to pre-populate interactive content e.g. faq's or contact pages. However using TensorflowJS in the browser comes with its own unique set of challenges and gotcha's.
Talk 2: From the Chinese Room Argument to the Church-Turing thesis - Dr Dean Petters
Searle’s Chinese Room thought experiment incorporates a number of assumptions about the role and nature of programs within the computational theory of mind. Two assumptions are analysed in this paper. One is concerned with how interactive we should expect programs to be for a complex cognitive system to be interpreted as having understanding about its environment and its own inner processes. The second is about how self-reflective programs might analyse their own processes. In particular, how self-reflection,and a high level of interactivity with the environment and other intelligent agents in the environment, may give rise to understanding in artificial cognitive systems. A further contribution that this pa-per makes is to demonstrate that the Church-Turing Thesis does not apply to interactive systems, and to self-reflective systems that in-corporate interactivity. This is an important finding because it means that claims about interactive and self-reflective systems need to be considered on a case by case basis rather than using lessons from relatively simple non-interactive and non-reflective computational models to generalise to all computational processes.
Dean D Petters Is a Senior Lecturer in Psychology at the University of Wolverhampton. His major research interest is using autonomous agent and multi-agent simulations to conduct research in Attachment Theory. In particular, using simulations to model the development of patterns of attachment during the first year of life. He also conducts empirical research on close adult relationships. Other major research foci include the study visual object recognition; 4e cognition; and the computational foundations of cognitive science. In March 2019 Dean was awarded the Bowlby-Ainsworth award for research in Attachment Theory. For "_explorations into the history, requirements, and prospects for computational modelling of human attachment_".