AAAI/EAAI Patrick Henry Winston Outstanding Educator Award: T(w)eaching AI in the Age of LLMs

Ultrafast laser surface structuring presents a compelling avenue for finely tuning material surface properties, such as coloration, wettability, tribological properties, etc. While nanoripples, the conventional laser-induced periodic surface structures, offer control over surface coloration, their effectiveness is limited to a narrow viewing angle. Addressing this limitation requires the development of techniques for generating alternative surface structures like nanobumps. Current methods for generating nanobumps, such as the double-pulse technique, are characterized by their complexity and high costs, posing significant challenges to widespread implementation. In this investigation, we introduce a novel approach for creating periodic nanobump structures through a straightforward double-scan method. This method involves subjecting the same surface to two consecutive laser scans with differing polarization directions.

Our study delves into the formation dynamics of these intriguing nanostructures, exploring the influence of laser parameters such as peak fluence, overlapping ratio, and polarization angle between scans. We propose and validate a formation mechanism to elucidate the observed phenomena. The resultant nanobump structures exhibit superior optical characteristics compared to nanoripples, particularly in terms of enhanced visibility across multiple viewing angles. By broadening the scope of achievable surface structures, our findings significantly expand the potential applications of ultrafast laser surface structuring. This research not only advances our understanding of laser-material interactions but also paves the way for the development of innovative surface engineering strategies with implications across diverse fields.

A Double-Scan Technique for Femtosecond Laser-Induced Nanobumps

Mechanistic interpretability is one of the most exciting and important research programs in current AI. My aim is to build some philosophical foundations for the program, along with setting out some concrete challenges and assessing progress to date. I will argue for the importance of propositional interpretability, which involves interpreting a system’s mechanisms and behavior in terms of propositional attitudes: attitudes (such as belief, desire, or subjective probability) to propositions (e.g. the proposition that it is hot outside). Propositional attitudes are the central way that we interpret and explain human beings and they are likely to be central in AI too. A central challenge is what I call thought logging: creating systems log all of the relevant propositional attitudes in an AI system over time. I will examine currently popular methods of interpretability (such as probing, sparse auto-encoders, and chain of thought methods) as well as philosophical methods of interpretation (including psychosemantics and representation theorems) to assess their strengths and weaknesses as methods of propositional interpretability.

Propositional Interpretability in Humans and AI Systems

Can AI Benefit Humanity?

Spatial reasoning is a core component of an agent’s ability to operate in or reason about the physical world. LLMs are widely promoted as having abilities to reason about a wide variety of domains, including commonsense. In this talk I will discuss the ability of state-of-the-art LLMs to perform commonsense reasoning, particularly with regard to spatial information. Across a wide range of LLMs, although they show abilities rather better than chance, they still struggle with many questions and tasks, for example when reasoning about directions, or topological relations.

Can Large Language Models Reason about Spatial Information?

In this keynote, Andrew will explore how current technologies, particularly agentic workflow, are revolutionizing the development of AI products. He will also delve into the role of AI in prototyping, showcasing how AI lowers the cost of software development and accelerates the prototyping process.

AI, Agents and Applications

This talk focuses on the use of foundation models to study problems in labor economics, in particular problems relating to the progression of worker careers and wages. In Vafa et al. (2024a), we introduced a transformer-based predictive model, CAREER, that predicts a worker’s next job as a function of career history (an “occupation model”). CAREER was initially estimated (“pre-trained”) using a large, unrepresentative resume dataset, which served as a “foundation model,” and parameter estimation was continued (“fine-tuned”) using data from a representative survey. CAREER had better predictive performance than benchmarks. Athey et al (2024) considers an alternative where the resume-based foundation model is replaced by a large language model (LLM). We convert tabular data from the survey into text files that resemble resumes and fine-tune the LLMs using these text files with the objective to predict the next token (word). The resulting fine-tuned LLM is used as an input to an occupation model. Its predictive performance surpasses all prior models. We demonstrate the value of fine-tuning and further show that by adding more career data from a different population, fine-tuning smaller LLMs surpasses the performance of fine-tuning larger models. In Vafa et al. (2024b), we apply and adapt this framework to the problem of gender wage decompositions, which require estimating the portion of the gender wage gap explained by career histories of workers. Classical methods for decomposing the wage gap employ simple linear models, and the resulting decompositions thus suffer from omitted variable bias (OVB), where covariates that are correlated with both gender and wages are not included in the model. We explore an alternative methodology for wage gap decomposition that employs CAREER as a foundation model. We prove that the way foundation models are usually trained might still lead to OVB, but develop fine-tuning algorithms that empirically mitigate this issue. We first provide a novel set of conditions under which an estimator of the wage gap based on a fine-tuned foundation model is root-n-consistent. Building on the theory, we then propose methods for fine-tuning foundation models that minimize OVB. Using data from the Panel Study of Income Dynamics, we find that history explains more of the gender wage gap than standard econometric models can measure, and we identify elements of history that are important for reducing OVB.

Predicting Career Transitions and Estimating Wage Disparities Using Foundation Models

The convergence of technologies—from foundation AI models to diverse sensors and actuators to ubiquitous connectivity—is transforming the nature of interactions in the physical and digital world. People have accelerated their collaborative connections and productivity through digital and immersive technologies, no longer limited by geography or language or access. Humans have also leveraged and interacted with AI in many different forms, with the advent of hyperscale AI models (i.e., large language models) forever changing (and at an ever-astonishing pace) the nature of human-AI teams, realized in this era of the AI “copilot.” Similarly, robotics and automation technologies now afford greater opportunities to work with and/or near humans, allowing for increasingly collaborative physical robots to dramatically impact real-world activities. It is the compounding effect of enabling all three capabilities, each complementary to one another in valuable ways, and we envision the triad formed by human-robot-AI teams as revolutionizing the future of society, the economy, and of technology.

Catalyzing the Future of Collaborative Human-Robot-AI Teaming

Power in the modern world - military, economic, geopolitical - is built on a foundation of computer chips. America has maintained its lead as a superpower because it has dominated advances in computer chips and all the technology that chips have enabled. (Virtually everything runs on chips: cars, phones, the stock market, even the electric grid.) Now that edge is in danger of slipping, undermined by the naive assumption that globalising the chip industry and letting players in Taiwan, Korea and Europe take over manufacturing serves America`s interests. Currently, as Chip War reveals, China, which spends more on chips than any other product, is pouring billions into a chip-building Manhattan Project to catch up to the US.

In Chip War economic historian Chris Miller recounts the fascinating sequence of events that led to the United States perfecting chip design, and how faster chips helped defeat the Soviet Union (by rendering the Russians’ arsenal of precision-guided weapons obsolete). The battle to control this industry will shape our future. China spends more money importing chips than buying oil, and they are China`s greatest external vulnerability as they are fundamentally reliant on foreign chips. But with 37 per cent of the global supply of chips being made in Taiwan, within easy range of Chinese missiles, the West's fear is that a solution may be close at hand. 

From "Chips War" to "Chiplets War"

Presidential Address

Test of Time Awards

Large language models (LLMs) have shown increasing capacity at planning and executing a high-level goal in a live computer environment (e.g. MiniWoB++). To perform a task, recent works often require a model to learn from trace examples of the task via either supervised learning or few/many-shot prompting. Without these trace examples, it remains a challenge how an agent can autonomously learn and improve its control on a computer, which limits the ability of an agent to perform a new task. We approach this problem with a zero-shot agent that requires no given expert traces. Our agent plans for executable actions on a partially observed environment, and iteratively progresses a task by identifying and learning from its mistakes via self-reflection and structured thought management. On the easy tasks of MiniWoB++, we show that our zero-shot agent often outperforms recent SoTAs, with more efficient reasoning. For tasks with more complexity, our reflective agent performs on par with prior best models, even though previous works had the advantages of accessing expert traces or additional screen information.

Underline digital video library

AAAI/EAAI Patrick Henry Winston Outstanding Educator Award: T(w)eaching AI in the Age of LLMs

A Double-Scan Technique for Femtosecond Laser-Induced Nanobumps

Propositional Interpretability in Humans and AI Systems

Can AI Benefit Humanity?

Can Large Language Models Reason about Spatial Information?

AI, Agents and Applications

Predicting Career Transitions and Estimating Wage Disparities Using Foundation Models

Catalyzing the Future of Collaborative Human-Robot-AI Teaming

From "Chips War" to "Chiplets War"

Presidential Address

Test of Time Awards

A Zero-Shot Language Agent for Computer Control with Structured Reflection

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES