United States

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.

AAAI 2025

Propositional Interpretability in Humans and AI Systems

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2022

invited talk

We are pleased to announce the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI-25), which will be held in Philadelphia, Pennsylvania at the Pennsylvania Convention Center from February 25 to March 4, 2025.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

### [Invited Speakers](https://aaai.org/conference/aaai/aaai-25/aaai-25-invited-speakers/)

Register [here](https://aaai.org/conference/aaai/aaai-25/registration/)

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.



We provide improved upper and lower bounds for the Min-Sum-Radii (MSR) and Min-Sum-Diameters (MSD) clustering problems with a bounded number of clusters $k$. In particular, we propose an exact MSD algorithm with a running-time of $n^{O(k)}$. We also provide $(1+\epsilon)$ approximation algorithms for MSR with a running-time of $O(kn) +(1/\epsilon)^{O(dk)}$ in metrics spaces of doubling dimension $d$. For MSD we get a similar approximation bound (with an extra $2^{O(k/\epsilon)}$ factor in the second term).

Our algorithms extend to $\alpha$-MSR, where radii are raised to an $\alpha$ power, for $\alpha>1$, whereas for $\alpha$-MSD we prove an exponential time ETH-based lower bound for $\alpha>\log 3$. All algorithms can also be modified to handle outliers. 

Moreover, we can extend the results to variants  that observe \emph{fairness} constraints, as well as to the general framework of \emph{mergeable} clustering, which includes many other popular clustering variants. We complement these upper bounds with ETH-based lower bounds for these problems, in particular proving that $n^{O(k)}$ time is tight for MSR and $\alpha$-MSR even in doubling spaces, and that $2^{o(k)}$ bounds are impossible for MSD.

Improved Fixed-Parameter Bounds for Min-Sum-Radii and Diameters k-Clustering and Their Fair Variants

Large Language Models (LLMs) have demonstrated remarkable proficiency in generating code. However, the misuse of LLM-generated (synthetic) code has raised concerns in both educational and industrial contexts, underscoring the urgent need for synthetic code detectors. Existing methods for detecting synthetic content are primarily designed for general text and struggle with code due to the unique grammatical structure of programming languages and the presence of numerous ``low-entropy'' tokens. Building on this, our work proposes a novel zero-shot synthetic code detector based on the similarity between the original code and its LLM-rewritten variants. Our method is based on the observation that differences between LLM-rewritten and original code tend to be smaller when the original code is synthetic. We utilize self-supervised contrastive learning to train a code similarity model and evaluate our approach on two synthetic code detection benchmarks. Our results demonstrate a significant improvement over existing SOTA synthetic content detectors, with AUROC scores increasing by 20.5\% on the APPS benchmark and 29.1\% on the MBPP benchmark.

Uncovering LLM-Generated Code: A Zero-Shot Synthetic Code Detector via Code Rewriting

Deep learning models have demonstrated exceptional performance in a variety of real-world applications. These successes are often attributed to strong base models that can generalize to novel tasks with limited supporting data while keeping prior knowledge intact. However, these impressive results are based on the availability of a large amount of high-quality data, which is often lacking in specialized biomedical applications. In such fields, models are often developed with limited data that arrive incrementally with novel categories. This requires the model to adapt to new information while preserving existing knowledge. Few-Shot Class-Incremental Learning (FSCIL) methods offer a promising approach to addressing these challenges, but they also depend on strong base models that face the same aforementioned limitations. To overcome these constraints, we propose AnchorInv following the straightforward and efficient buffer-replay strategy. Instead of selecting and storing raw data, AnchorInv generates synthetic samples guided by anchor points in the feature space. This approach protects privacy and regularizes the model for adaptation. When evaluated on three public physiological time series datasets, AnchorInv exhibits efficient knowledge forgetting prevention and improved adaptation to novel classes, surpassing state-of-the-art baselines.

AnchorInv: Few-Shot Class-Incremental Learning of Physiological Signals via Feature Space-Guided Inversion

Building a universal multilingual automatic speech recognition (ASR) model that performs equitably across languages has long been a challenge due to its inherent difficulties.
To address this task we introduce a Language-Agnostic Multilingual ASR pipeline through orthography Unification and language-specific Transliteration (LAMA-UT).
LAMA-UT operates without any language-specific modules while matching the performance of state-of-the-art models trained on a minimal amount of data.
Our pipeline consists of two key steps.
First, we utilize a universal transcription generator to unify orthographic features into Romanized form and capture common phonetic characteristics across diverse languages. 
Second, we utilize a universal converter to transform these universal transcriptions into language-specific ones.
In experiments, we demonstrate the effectiveness of our proposed method leveraging universal transcriptions for massively multilingual ASR.
Our pipeline achieves a relative error reduction rate of 45\% when compared to Whisper and performs comparably to MMS, despite being trained on only 0.1\% of Whisper's training data.
Furthermore, our pipeline does not rely on any language-specific modules.
However, it performs on par with zero-shot ASR approaches which utilize additional language-specific lexicons and language models.
We expect this framework to serve as a cornerstone for flexible multilingual ASR systems that are generalizable even to unseen languages.

LAMA-UT: Language Agnostic Multilingual ASR Through Orthography Unification and Language-Specific Transliteration

Surface-from-gradients (SfG) aims to recover a three-dimensional (3D) surface from its gradients. Traditional methods encounter significant challenges in achieving high accuracy and handling high-resolution inputs, particularly facing the complex nature of discontinuities and the inefficiencies associated with large-scale linear solvers. Although recent advances in deep learning, such as photometric stereo, have enhanced normal estimation accuracy, they do not fully address the intricacies of gradient-based surface reconstruction. To overcome these limitations, we propose a Fourier neural operator-based Numerical Integration Network (FNIN) within a two-stage optimization framework. In the first stage, our approach employs an iterative architecture for numerical integration, harnessing an advanced Fourier neural operator to approximate the solution operator in Fourier space. Additionally, a self-learning attention mechanism is incorporated to effectively detect and handle discontinuities. In the second stage, we refine the surface reconstruction by formulating a weighted least squares problem, addressing the identified discontinuities rationally. Extensive experiments demonstrate that our method achieves significant improvements in both accuracy and efficiency compared to current state-of-the-art solvers. This is particularly evident in handling high-resolution images with complex data, achieving error fewer than 0.1 mm on tested objects.

FNIN: A Fourier Neural Operator-based Numerical Integration Network for Surface-from-gradients

Dependency-aware spatial crowdsourcing (DASC) addresses the unique challenges posed by subtask dependencies in spatial task assignment. This paper investigates the task assignment problem in DASC and proposes a two-stage Recommend and Match Optimization (RMO) framework, leveraging multi-agent reinforcement learning for subtask recommendation and a multi-dimensional utility function for subtask matching. The RMO framework primarily addresses two key challenges: credit assignment for subtasks with interdependencies and maintaining overall coherence between subtask recommendation and matching. Specifically, we employ meta-gradients to construct auxiliary policies and establish a gradient connection between two stages, which can effectively address credit assignment and joint optimization of subtask recommendation and matching, while concurrently accelerating network training. We further establish a unified gradient descent process through gradient synchronization across recommendation networks, auxiliary policies, and the matching utility evaluation function. Experiments on two real-world datasets validate the effectiveness and feasibility of our proposed approach.

Gradient-Guided Credit Assignment and Joint Optimization for Dependency-Aware Spatial Crowdsourcing

Offline safe reinforcement learning (RL) has emerged as a promising approach for learning safe behaviors without engaging in risky online interactions with the environment. Most existing methods in offline safe RL rely on cost constraints at each time step (derived from global cost constraints) and this can result in either overly conservative policies or violation of safety constraints. In this paper, we propose
to learn a policy that generates desirable trajectories and avoids undesirable trajectories. To be specific, we first partition the pre-collected dataset of state-action trajectories into desirable and undesirable subsets. Intuitively, the desirable set contains high reward and safe trajectories, and undesirable set contains unsafe trajectories and low-reward safe trajectories. Second, we learn a policy that generates desirable trajectories and avoids undesirable trajectories, where (un)desirability scores are provided by a classifier learnt from the dataset of desirable and undesirable trajectories. This approach bypasses the computational complexity and stability issues of a min-max objective that is employed in existing methods. Theoretically, we also show our approach’s strong connections to existing learning paradigms involving human feedback. Finally, we extensively evaluate our method using the DSRL benchmark for offline safe RL. Empirically, our method outperforms competitive baselines, achieving higher rewards and better constraint satisfaction across a wide variety of benchmark tasks.

Offline Safe Reinforcement Learning Using Trajectory Classification

Autonomous robots are becoming more versatile and widespread in our daily lives. From autonomous vehicles to companion robots for senior care, these human-centric systems must demonstrate a high degree of reliability in order to build trust and, ultimately, deliver social value. How safe is safe enough for robots to be wholeheartedly trusted by society? Is it sufficient if an autonomous vehicle can avoid hitting a fallen cyclist 99.9% of the time? What if this rate can only be achieved by the vehicle always stopping and waiting for the human to move out of the way? I argue that, for trustworthy deployment of robots in human-populated space, we need to complement standard statistical methods with clear-cut robust safety assurances under a vetted set of operation conditions. We need runtime learning to minimize the robot’s performance loss during safety-enforcing maneuvers by reducing its inherent uncertainty induced by its human peers, for example, their intent (does a human driver want to merge, cut behind, or stay in the lane?) or response (if the robot comes closer, how will the human react?). We need to close the loop between the robot’s learning and decision-making so that it can optimize efficiency by anticipating how its ongoing interaction with the human may affect the evolving uncertainty, and ultimately, its long-term performance.

From Gambits to Assurances: Game-Theoretic Integration of Safety and Learning for Interactive Robotics

My thesis primarily focuses on hyper-spectral image generation from frequency spectrums for downstream computer vision tasks. Hyper-spectral images are images with more than three channels commonly created by special hyper-spectral cameras or from frequency spectrums of various sensing applications such as radargrams or distributed acoustic sensing (DAS) systems. The range of frequencies considered in a frequency spectrum is typically too large to map one frequency to one image channel, i.e. we generally consider a frequency spectrum of 2500 Hz. Frequencies need to be binned together in frequency bands where each band forms one image channel. Usually, frequency bands are created either by expert knowledge or trial-and-error.
I research how filters can be trained to automatically select frequencies and bin them into frequency bands. My aim is to represent a variety of signal information and decrease noise. Signal representation is optimised for object detection on time-sequenced images with a set number of image channels. The object detection task consists of localising and classifying events in the generated hyper-spectral images. Events are typically types of intrusions, structural changes, or defined actions and structures, e.g. someone climbing a fence. Events and noise often share at least some frequencies and vary between application types.

Downloads

Next from AAAI 2025

Improved Fixed-Parameter Bounds for Min-Sum-Radii and Diameters k-Clustering and Their Fair Variants

Similar lecture

Tractable Explanations for d-DNNF Classifiers

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2025

Improved Fixed-Parameter Bounds for Min-Sum-Radii and Diameters k-Clustering and Their Fair Variants

Similar lecture

Tractable Explanations for d-DNNF Classifiers

Downloads