MSV-2035 Astronomy Document - Inside Design - FINAL - FINAL

Astronomy & Astrophysics 111 The usage of AI/ML techniques requires access to high performance computing and in the coming years there is a pressing need to combine data archives and virtual observatories with super computing systems so that very large data sets can be analysed at source and users do not have to transfer raw data or work with raw data on their computers. Much of the development in the application of AI/ML in astronomy is aided by developments outside astronomy, particularly in the field of computer science. We are seeing rapid developments on many fronts such as convolutional neural networks utilising deep learning for image classification, generative pre-trained transformers for text generation, and generative adversarial networks for text-to-speech, text-to-image and text-to-video applications. Ef- forts are on to develop a generalised AI, a system capable of performing a wide range of tasks at a human-level intelligence. These developments need to be followed, and such emerging technologies need to be adopted for research problems in astronomy. The large volume data sets available in astronomy will, in turn, aid these developments. There is considerable expertise available in the industry in the area of big data andAI/ML, that can be capitalised and harnessed by the research sector through systematic efforts. MEGA SCIENCE VISION-2035