Annotated slide deck for lightning talk “AI at Kew - motivation and context” from the AI for the Natural Sciences symposium held on 3rd November 2022 at the Natural History Museum, London, UK.
An overview of how we work with AI at Kew - given by Nicky Nicolson (NN) with colleagues also at the event and available for discussion.
Personal context: NN transitioned from software development into research, and is interested in how software development practices (reuse, automation, version control, dependency management, continuous integration) can be used in research.
Thinking about how we work - “AI” could also also mean:
- Accessible information: we’re digitising our collections (Image credit: RBG Kew)
- Authoritative interactions: how experts use and interpret that data: using digital technologies to accelerate the process of taxonomy (Image credit: RBG Kew)
- Adaptation and innovation: reusing tools & techniques, facilitating Open Science skills development. Forming partnerships with different disciplines (Image credit: Tony Iwane)
The work to define and build the digital extended specimen can advance using a lot of different techniques, learning as we go.
We can contribute to large scale machine learning approaches, like the clustering and specimen duplicate detection run on the GBIF data portal…
… to hands-on tools for researchers. These toolkits can allow them to generate data for machine learning applications in-context, as they conduct their research, and to verify the outputs of more automated approaches. echinopscis.github.io
We’re also working to develop representations of specimens that allow researchers to interact with the large scale image data in different ways. Walker BE, Tucker A, Nicolson N. Harnessing Large-Scale Herbarium Image Datasets Through Representation Learning. Front Plant Sci. 2022 doi:10.3389/fpls.2021.806407
… and we’re exploring some of the patterns of resource allocation, looking at type citations and if the institutions into which types specimens are lodged are changing over time. This is our first link-up with a digitial humanities research group.