Election Integrity with AI & CNNs for Atomic Imaging

The MLAI Meetup is a community for AI researchers and professionals which hosts monthly talks on exciting research. Our format is:

• 6:00 - 6:20: Socializing
• 6:20 - 6:30: Announcements and AI news
• 6:30 - 7:40: Talk(s) and Q&A
• 7:40 - 8:00 Networking
• 8:00: Head to the nearest pub for dinner

Election Integrity: Using AI to Audit Election Outcomes

Presenter: Alexander Ek

Abstract: Elections are the cornerstone of democratic societies, and ensuring their integrity and trustworthiness is crucial. In this talk, we explore cutting-edge techniques that combine AI, statistical analysis, machine learning, and combinatorial optimisation to audit election results. This centres around risk-limiting audits, which is a state-of-the-art procedure for checking and correcting election results.
Unfortunately, elections like those in Australia, which allows voters to rank candidates, are notoriously hard to audit due to the factorial amount of possibilities. With the help of AI, however, recent developments have been able to overcome this hurdle. Join us as we demystify a key component behind creating trustworthy elections in Australia and around the world.

Speaker Bio: Alexander Ek is a research fellow at the Monash University and an associate investigator at OPTIMA. In 2022, he received his PhD in Computer Science, also from Monash University, with a focus on combinatorial optimisation under uncertainty. Since, his focus has been on election integrity and auditing, combining statistics, machine learning, AI, and combinatorial optimisation. His other interests include algorithmics, uncertainty and risk, and applied mathematics.

Enhanced Automation of Scanning Probe Microscopy through Model Reprogramming

Presenter: Scott Bennett

Abstract: Scanning Probe Microscopy (SPM) encompasses a suite of techniques used for imaging and manipulating matter at the nanoscale. Automating these tools has recently become a focus due to their labor-intensive and challenging nature. Emerging research has demonstrated the potential of Deep / Reinforcement Learning to automate these systems successfully. However, the process of automation remains largely inaccessible to the average SPM user due to the quantity of data required and poor generalisation. Our research aims to develop a framework which leverages Adversarial Model Reprogramming and synthetic data generation to create models that are more generalised with less data - hence making this automation more accessible.

Speaker Bio: Scott Bennett is currently completing his honours research at Monash University in collaboration with Monash's Surface Nanophysics and AI Research Groups. He has a variety of experience working in experimental physics and artificial intelligence, having previously led the Monash AI Student Research Team, DeepNeuron. Other interests include: AI ethics, neuromorphic computing, entrepreneurship and computer vision.