Yasemin Erden on BBC

AISB Committee member, and Philosophy Programme Director and Lecturer, Dr Yasemin J. Erden interviewed for the BBC on 29 October 2013. Speaking on the Today programme for BBC Radio 4, as well as the Business Report for BBC world N...


Mark Bishop on BBC ...

Mark Bishop, Chair of the Study of Artificial Intelligence and the Simulation of Behaviour, appeared on Newsnight to discuss the ethics of ‘killer robots’. He was approached to give his view on a report raising questions on the et...


AISB YouTube Channel

The AISB has launched a YouTube channel: http://www.youtube.com/user/AISBTube (http://www.youtube.com/user/AISBTube). The channel currently holds a number of videos from the AISB 2010 Convention. Videos include the AISB round t...


Lighthill Debates

The Lighthill debates from 1973 are now available on YouTube. You need to a flashplayer enabled browser to view this YouTube video  



AISB opportunities Bulletin Item

PhD position available, University of Plymouth: Errors and biases in machine visual identification of plankton

PhD position available
University of Plymouth

Errors and biases in machine visual identification of plankton

A new PhD position is available, closing date 18th Sept. The work is with Dr Phil Culverhouse, in the Natural Object Categoristion Group within the Centre for Robotics & Neural Systems at the University of Plymouth, UK.

Title: Errors and biases in machine visual identification of plankton

Automated visual identification of marine plankton software requires training data to initialise class characteristics and establish class boundaries between groups to be identified. This is normally achieved by establishing image sets for each class, selected by experts in plankton identification (taxonomists and marine ecologists). However, these experts are biased and make mistakes, which manifest as images not in-class being labelled as in-class. In addition, for many microplankton species experts do not yet agree on the classification. This is compounded by uncertainty as to classification, resulting in probability assignments to the correct category ie. Image 1 could be species A (80% confidence) or species B (20% confidence).   It is suspected that the nature of the regimes used to train machine classifiers reduces the impact of these human errors. But no robust experimental or theoretical arguments have been presented to support this.
  The PhD programme will research the theoretical underpinnings of multi-expert consensus and the effects of errors and biases on the performance of machine vision-based categorisation tools, establishing the theoretical error limits for all machine visual object identification systems, with special relevance to marine plankton identification.

Culverhouse is co-chair of the SCOR WG130 working group on Automated Visual Identification of Plankton. This allows access to experts and data sets from marine scientists across the world.

The work is in collaboration with SAFHOS and PML. The position is fully funded available for 3 years.

Further information is available from p.culverhouse@plymouth.ac.uk

Application packs can be obtained from c.watson@plymouth.ac.uk