This event is held in collaboration between the Research Software Engineering (RSE) group and the Centre for Machine Intelligence (CMI) at the University of Sheffield, and the N8 CIR.
The afternoon will consist of talks and walkthroughs on best practices for research, design, development, and deployment of AI. It will focus on practical aspects such as tooling, optimisation, profiling, tips and tricks to supercharge AI in your research!
In-person spaces are limited; please register to attend either in-person or remotely.
AGENDA
12:00-12:05 Welcome
12:05-13:00 Community lightning talk & Networking Lunch
Speakers from the community
Lightning talks from members of the community, followed by buffet lunch, soft drinks and coffee.
13:00-14:00 Talk by Nvidia (Title TBC)
Overview of the current AI landscape and deep dive into a topic (TBC).
14:00-14:30 Responsible use of (Generative) AI in research and innovation
Denis Newman-Griffis, University of Sheffield
The general-purpose use of AI technologies in research and innovation is rapidly evolving. The University of Sheffield has recently developed a set of Principles for Using GenAI in Research and Innovation, which aim to provide a starting point for students and staff at the university exploring the use of GenAI in their own research. This talk will outline the thinking behind the University Principles and how they might be applied in research practice, and will highlight current developments around AI use in the wider national and international research systems.
14:30-15:00 Creating your own agent with LangChain
Shaun Donnelly, University of Sheffield
A walkthrough covering how to create an agent, equip it with tools, and demonstrate their use, including multiple tool interactions. The session also explains how the agent decides which tools to use and when, illustrating the decision-making process behind effective tool-driven agents.
15:00-15:15 Coffee break
15:15-15:45 Table talks (topic TBC)
15:45-16:15 Evaluating AI tools for finding research studies for evidence synthesis
Su Golder, University of York
We will present the results of an evaluation on the use of the AI tool Elicit to identify relevant research studies for systematic reviews. We pay particular attention to the methods used and challenges along the way, using four real-world case study reviews.
16:15-16:45 The benefits (and some dangers) of data visualization for explainable AI
Roy Ruddle, University of Leeds
To use visualization effectively in explainable AI (XAI), you first need to identify: (a) the XAI tasks you are performing, (b) specific questions you want to answer, and (c) the data types that are involved. In this talk, I will illustrate how different visualization techniques map to a wide range of XAI tasks and questions. Some of the visualization techniques are well-known (e.g., scatterplots, bar charts, and heat maps), but others are more unusual (e.g., violin plots, beeswarm plots, and parallel coordinates). I will also show you how to avoid being misled by visualizations and some common bloopers.
16:45-17:00 Wrap-up and feedback