Commoditising the use of AI for scientists: how AI-agents assist and transform scientific discovery

AI is revolutionising science, offering immense potential to tackle complex scientific challenges. CSIRO has recently initiated a strategic research project called Science Digital. The aim of this project is to develop an LLM (Large Language Models) agent platform to assist domain scientists who have limited expertise in AI and software engineering in creating and orchestrating responsible, personalised, proactive, and context-aware agents to help them achieve their research goals more effectively and efficiently. In this talk, we will first provide an overview of the agent platform architecture, highlighting how various design patterns and tools are leveraged to build this platform. Then, we dive into some of the agents that are specifically designed to assist domain experts in their daily jobs to sift through a large amount of ever-growing scientific literature which relies on identifying relevant content, reviewing, and summarising scientific knowledge, as well as distilling fine-grained information from the literature that could assist with scientific discovery. These agents are created based on Natural language processing and information retrieval techniques, as well as use-case specific evaluation frameworks to ensure their suitability. This session is intended for scientists that are interested in tools for accelerating scientific discovery. We want to inspire Scientists to think about how they could apply AI and to share proposed tools that could elevate their skill level in applying AI. The session will be divided into two main parts, starting with the speakers introducing the Science Digital Project and then an interactive session where the audience can ask questions.