“How can ChatGPT help your research, a discussion workshop”
The School of Chemical and Physical Sciences had a lunchtime workshop on how we are using ChatGPT and thoughts to consider into the future. It was a packed room showing the wide interest in this topic. Luke Liu from Chemistry told us about how he’s been using ChatGPT as a “non-coder” to write python scripts. This has enabled him, for example, to quickly produce a plotting GUI which he could compile and copy onto a different computer to enable him to easily plot some research data.
Andre Geldenhuis from Research Computing spoke about how to use ChatGPT to create and understand sbatch run scripts for Rāpoi. This was a great example of how to use ChatGPT for a fairly simple problem, but one that has lots of different options that could take you a long time to read through and understand. This could potentially be a big timesaver the next time you want to do something slightly different from the standard like selecting a particular node to run on.
Geoffrey Weal, a post-doc in Chemistry spoke about progress in mining scientific literature using ChatGPT (or other natural language processing algorithms). In his area of research, there is a lot of literature going back many years that may contain data tables in non-machine-readable format, or embedded in the text of a paper. He and his group are working on ways to in the first instance select useful papers without a researcher’s input, and eventually automatically extract these data.
Baptiste Auguie from Physics, in his talk ‘Behold: The Three Lows of Robotics’ provided an important balancing argument by summarising various reasons one might want to refrain from using (enthusiastically) ChatGPT. He mentioned some of the IP issues around the training sets that are being used to train the network (i.e. all of your code on GitHub), and the concerns around the human rights of people in less developed countries being used to ensure general content from the internet used for training is clean, as well as more philosophical arguments about the future of AI.
