Professional Documents
Culture Documents
Macleod Et Al 2010 Nature
Macleod Et Al 2010 Nature
OPINION
A
n imaging system designed to iden- virtually every other scientific discipline. animal group about 65.5 million years ago1
LefT To righT: NaT. hisT. Mus., LoNdoN; M. BeNfieLd/guLf serPeNT ProjecT; N. caTTLiN/fLPa
tify marine zooplankton was recently Automating species identification using resulted in species lists that were so differ-
adopted by scientists working for the technologies developed by researchers in ent as to make consensus impossible. Such
US government to monitor the Deepwater pattern recognition, artificial intelligence inconsistencies shouldn’t be a surprise given
Horizon oil spill. By measuring the size of oil and machine learning would transform alpha that, in controlled visual-cognition studies,
droplets produced after chemical dispersants taxonomy from a cottage industry depend- humans frequently miss items presented in
had broken up the oil, modellers could predict ent on the expertise of a few individuals to a scene, count some objects more than once
the depths at which the plume was accumu- a testable and verifiable science accessible to and misclassify others.
lating. Only two instruments exist that can anyone needing to recognize objects. Indeed, Hopes are high among researchers and
measure oil droplets while distinguishing a concerted interdisciplinary research and funding bodies that DNA bar-coding, by
them from other matter suspended in the development effort, within the next decade, which a species is recognized according to
water column, such as zooplankton, marine could lead to automated systems capable of a marker in its mitochondrial genome, will
snow and gas bubbles at depths of down to high-throughput identifications for hundreds increase the accuracy of identifications — and
1,500 metres. The deployment, by the US or thousands of categories of living as well as ease bottlenecks resulting from a shortage of
National Oceanic and Atmospheric Admin- non-living specimens. trained and experienced taxonomists. But bar
istration, of one — the digital holographic codes are generally used to assign organisms
imaging (DHI) system, developed jointly by Human error to taxonomic categories that have already
the Massachusetts Institute of Technology Many taxonomists use sophisticated tech- been defined on the basis of morphological
in Cambridge and the Woods Hole Oceano- nologies to capture images, sounds, even the traits. In other words, a bar code isn’t useful
graphic Institution in Massachusetts — is a smells and tastes of biological specimens. until the reference species has already been
working example of something that should But most routine identifica- identified by multiple experts.
be happening on a grand scale: the shared use tions involve a small group
across diverse disciplines of generalized auto- of experts scattered around
“Humans miss items, The technique is still relatively
expensive, slow and difficult
mated identification technologies. the world assessing diag- count some objects to implement in the field
Taxonomists who identify, describe and nostic data qualitatively — more than once and except in certain situations
name species (who practise alpha taxonomy, commonly the size, shape or — for example in laboratories
as it is known in the trade) are central to many texture of specimens, or the
misclassify others.” on oceanographic research
research programmes in applied biology, presence or absence of cer- vessels. Moreover, research-
ecology and conservation. University cuts tain features. Surprisingly few blind-test stud- ers frequently need to identify non-living
are shrinking this already small community. ies have been published to assess the accuracy objects as well as living ones. Ecologists study-
What’s more, there is no tradition of — much of taxonomists’ findings objectively1–7. Those ing plankton, for example, commonly count
less a requirement for — independent testing that have been carried out are worrying. For ‘fibres’, ‘detritus’ or ‘egg-like particles’ that may
and verification of the accuracy of the identi- instance, a blind test to resolve a controversy or may not be alive.
fications that taxonomists produce, unlike in about the pattern of extinction in one marine In focusing on bar-coding, stakeholders have
154
© 2010 Macmillan Publishers Limited. All rights reserved
NATURE|Vol 467|9 September 2010 OPINION
overlooked the greater promise of machine- of which may not be detectable by humans — within a standard set of complex but general-
learning to transform taxonomy and the into identification programs. For example, bat ized scenes, such as in photographs of a coral
identification of natural objects in general. echolocation calls that are outside the range of reef, would attract public interest and encour-
the human auditory system. Finally, software age diverse groups of scientists to explore the
DAISY, DAISY, give me your answer, do designers are improving the user interfaces of technologies available. This would be similar to
Computer systems now exist for classifying classification programs. the ‘visual object classes’ challenge recently set
objects into between 2 and 30 categories. These Currently, grant applications for such up by the EU-funded Network of Excellence
systems already deliver faster, more accurate interdisciplinary projects are falling between on Pattern Analysis, Statistical Modelling and
and more consistent semi- or fully automated the boundaries defined by funding bodies in Computational Learning (PASCAL), which
identifications than any human taxonomist. engineering and the life sciences. Funding promotes the development of computer sys-
For instance, a group of entomologists at the specifically for collaborations on automated tems that recognize types of common objects.
Natural History Museum in London have species identification should be supplied by This investment would pay huge dividends
used the Digital Automated Identification the European Union’s framework programmes across a range of disciplines. In the past 50 years,
System (DAISY) to identify with 100% accu- (Europe’s main instrument for funding academic centres worldwide have cut back or
racy 15 species of parasitic wasp from digital research), and national research councils such discontinued many taxonomic training and
images of wings, with each identification as the US National Science Foundation and the research programmes. As a result, there are
taking less than a second8. Similarly, oceanog- UK Natural Environment Research Council. only about 4,000–6,000 professional taxono-
raphers from the University of Plymouth, UK, Charitable organizations, such as the Wellcome mists worldwide10, only a subset of whom are
have used the Dinoflagellate Categorisation by routinely engaged in species identifications.
155
© 2010 Macmillan Publishers Limited. All rights reserved