Agenda - UVA Conference on Leadership in Business, Data and Intelligence

With the rise of generative AI, there has been a massive increase in demand for compute.  There are increasing concerns about the ability to build the server farms necessary to generate this processing power in a timely fashion, largely due to the inability to secure electricity for new data centers. At the same time, there has been a fundamental shift in US electrical generation where most new capacity is coming from renewables creating hopes for the potential to decarbonize electrical generation while raising concerns about intermittency and the resilience of the grid.  In this talk, Professor Mike Lenox will discuss how these simultaneous forces, a massive increase in demand and a fundament shift in supply, are creating both opportunities and challenges that will prove disruptive to the energy sector and the sectors heavily reliant on electricity.  Lenox will advance propositions on how this will likely play out in terms of technology innovation, competitive structure, and emerging geopolitical positioning to dominate new technologies and secure valuable supply chains.

The widespread practice of indiscriminate data scraping to train and fine-tune generative models raises significant legal and ethical concerns, particularly regarding compliance with data protection laws such as the General Data Protection Regulation (GDPR), leading to unauthorized use of personal information, prompting growing debate within the academic and regulatory communities. In addition, companies face the challenge of maintaining data utility while ensuring compliance with stringent data protection regulations like GDPR. In this talk, Professors Chirag Agarwal and Sam Levy discuss different ways to operationalize the right to data privacy in the Ethical AI pipeline. This encompasses the lifecycle of ethical AI, from data curation and collection to training models with privacy-preserving techniques, post-processing models to enhance privacy, and evaluation mechanisms to measure compliance and performance. Their talk introduces methodologies like Privacy-Preserving Data Fusion (PPDF), which ensures user anonymity while enabling robust data fusion, and RegText, a framework for embedding imperceptible spurious correlations in datasets to render them unlearnable, protecting public data from exploitation. Looking forward, they envision a future where ethical AI pipelines are enhanced with advanced methodologies to operationalize data privacy, aligning AI development with evolving legal standards and ethical imperatives.

This session focuses on pathways to sustainable and responsible AI development. Michael Albert will explore the need for marketplaces for data in AI, emphasizing their role in fostering the creation of high-quality training data and ensuring fair compensation for data creators. Tom Hartvigsen will present on AI model editing techniques that maintain model relevance while mitigating bias. Together, these talks provide actionable insights for building AI systems that balance innovation, fairness, and long-term viability.

Earlier this year, UVA's Cavalier Autonomous Racing team, led by Prof. Madhur Behl, made history by setting a new world record at the Indianapolis Motor Speedway - becoming the first American team to win this global competition and reaching 184 mph autonomously at the racetrack. Autonomous racing is about pushing the boundaries of what’s possible in Robotics and AI. But why, despite such breakthroughs, have autonomous driving systems not met the high expectations set by many? What are the AI challenges that continue to hold us back?

In this talk, I will first describe this missing piece - physical intelligence. I will then show how high-speed autonomous racing provides a unique proving ground to test the limits of AI’s physical intelligence. Leveraging more than a decade of experience from high-speed autonomous racing, particularly with the full-scale Cavalier Autonomous Racing Indy car and the F1Tenth platform, I will demonstrate how racing at speeds exceeding 160 mph while in close quarters with other vehicles presents unique AI challenges. I will recount our journey from algorithms to accelerations, and the rigorous engineering required to develop an autonomous racing car from scratch. Despite progress, autonomous racing has yet to match expert racing drivers’ skills or navigate the chaos of dense, multi-car racing in the real world; indicating that several more laps are needed on our journey towards artificial general “driving” intelligence.

The first part of this session will be given by Professor Sarah Lebovitz and is entitled “AI in Action: Unpacking the Reality of Human-AI Collaboration in Healthcare”. This talk explores how AI is reshaping work practices, by zooming in on the context of healthcare and diagnostic AI for radiology. Lebovitz will discuss the field's enthusiastic embrace of AI, marked by rapid advancements in image recognition technology and growing institutional investment. However, through qualitative field research, Levovitz uncovered the complex realities of AI implementation inside hospitals, revealing critical challenges in human-AI collaboration and evaluating AI tools. By examining these dynamics, the talk reflects on broader implications for technology’s role in transforming work and equity in organizations.

The second part of the session is entitled “Harnessing LLMs for Psychological Attribute Extraction in Business Research” and will be given by Professor Reza Mousavi. In this presentation, Mousavi will briefly review four key (natural language processing) NLP paradigms that have evolved over the past 70 years: lexicon-based approaches, custom-built models, fine-tuned masked language models (FLMs), and large language models (LLMs). Using seven cases, Mousavi will benchmark and compare these methods based on their ability to extract psychological attributes from natural language. His findings show that LLMs emerge as the most reliable and fair approach, particularly in addressing biases related to gender and race. Additionally, he will cover various techniques—soft and hard prompting—and methods like agentic workflows and infusing psychological traits to achieve the highest accuracy from various LLMs (e.g., GPTs, Llamas). This work informs businesses and academic research across disciplines on building efficient pipelines for understanding humans based on language they use.

This will be a wide-ranging panel discussion about the key leadership skills needed to navigate an increasingly AI-savvy workforce, with insights on managing talent, fostering adaptability, and driving innovation in a rapidly evolving technological landscape. 

This panel brings together leading experts from industry to explore the evolving intersection of AI and talent management. Moderated by Dr. Mona Sloane, a sociologist studying AI's societal impact, the discussion will feature insights from Alan Susi (S&P Global's VP of Organizational Analytics), Ian O'Keefe (founder of iKona Analytics), and Dirk Petersen (VP of Insight222). The panel will examine increasingly important questions about AI's role in workplace transformation, including its impact on productivity, creativity, and job creation, while addressing the tensions between technological advancement and maintaining the human element in business. Drawing from their extensive experience in people analytics and organizational effectiveness, the panelists will discuss how leaders are navigating AI adoption decisions and organizations' responsibilities in balancing increased productivity with workforce sustainability.

How is the tech sector itself experiencing the promise and potential of AI? From startup to mid-size to Meta, panelists will explore the risk-innovation tension and the private-public sector dynamic in ethical development and use of the technology. 

Coming soon