Colour controlled image generation and manipulation are of interest to artists and graphic designers. Vector Quantised Variational AutoEncoders (VQ-VAEs) with autoregressive (AR) prior are able to produce high quality images, but lack an explicit representation mechanism to control colour attributes. We introduce DualVAE, a hybrid representation model that provides such control by learning disentangled representations for colour and geometry. The geometry is represented by an image intensity mapping that identifies structural features. The disentangled representation is obtained by two novel mechanisms: (i) a dual branch architecture that separates image colour attributes from geometric attributes, and (ii) a new ELBO that trains the combined colour and geometry representations. DualVAE can control the colour of generated images, and recolour existing images by transferring the colour latent representation obtained from an exemplar image. We demonstrate that DualVAE generates images with FID nearly two times better than VQ-GAN on a diverse collection of datasets, including animated faces, logos and artistic landscapes.

Coherent measurement of quantum signals used for continuous-variable (CV) quantum key distribution (QKD) across satellite-to-ground channels requires compensation of phase wavefront distortions caused by atmospheric turbulence. One compensation technique involves multiplexing classical reference pulses (RPs) and the quantum signal, with direct phase measurements on the RPs then used to modulate a real local oscillator (RLO) on the ground - a solution that also removes some known attacks on CV-QKD. However, this is a cumbersome task in practice - requiring substantial complexity in equipment requirements and deployment. As an alternative to this traditional practice, here we introduce a new method for estimating phase corrections for an RLO by using only intensity measurements from RPs as input to a convolutional neural network, mitigating completely the necessity to measure phase wavefronts directly. Conventional wisdom dictates such an approach would likely be fruitless. However, we show that the phase correction accuracy needed to provide for non-zero secure key rates through satellite-to-ground channels is achieved by our intensity-only measurements. Our work shows, for the first time, how artificial intelligence algorithms can replace phase-measuring equipment in the context of CV-QKD delivered from space, thereby delivering an alternate deployment paradigm for this global quantum-communication application.

The rapid advancement of artificial intelligence (AI) systems suggests that artificial general intelligence (AGI) systems may soon arrive. Many researchers are concerned that AIs and AGIs will harm humans via intentional misuse (AI-misuse) or through accidents (AI-accidents). In respect of AI-accidents, there is an increasing effort focused on developing algorithms and paradigms that ensure AI systems are aligned to what humans intend, e.g. AI systems that yield actions or recommendations that humans might judge as consistent with their intentions and goals. Here we argue that alignment to human intent is insufficient for safe AI systems and that preservation of long-term agency of humans may be a more robust standard, and one that needs to be separated explicitly and a priori during optimization. We argue that AI systems can reshape human intention and discuss the lack of biological and psychological mechanisms that protect humans from loss of agency. We provide the first formal definition of agency-preserving AI-human interactions which focuses on forward-looking agency evaluations and argue that AI systems - not humans - must be increasingly tasked with making these evaluations. We show how agency loss can occur in simple environments containing embedded agents that use temporal-difference learning to make action recommendations. Finally, we propose a new area of research called "agency foundations" and pose four initial topics designed to improve our understanding of agency in AI-human interactions: benevolent game theory, algorithmic foundations of human rights, mechanistic interpretability of agency representation in neural-networks and reinforcement learning from internal states.

Large language models (LLMs) have a wealth of knowledge that allows them to excel in various Natural Language Processing (NLP) tasks. Current research focuses on enhancing their performance within their existing knowledge. Despite their vast knowledge, LLMs are still limited by the amount of information they can accommodate and comprehend. Therefore, the ability to understand their own limitations on the unknows, referred to as self-knowledge, is of paramount importance. This study aims to evaluate LLMs' self-knowledge by assessing their ability to identify unanswerable or unknowable questions. We introduce an automated methodology to detect uncertainty in the responses of these models, providing a novel measure of their self-knowledge. We further introduce a unique dataset, SelfAware, consisting of unanswerable questions from five diverse categories and their answerable counterparts. Our extensive analysis, involving 20 LLMs including GPT-3, InstructGPT, and LLaMA, discovering an intrinsic capacity for self-knowledge within these models. Moreover, we demonstrate that in-context learning and instruction tuning can further enhance this self-knowledge. Despite this promising insight, our findings also highlight a considerable gap between the capabilities of these models and human proficiency in recognizing the limits of their knowledge.

Emergent properties have been widely adopted as a term to describe behavior not present in smaller models but observed in larger models. Recent work suggests that the trade-off incurred by quantization is also an emergent property, with sharp drops in performance in models over 6B parameters. In this work, we ask "are quantization cliffs in performance solely a factor of scale?" Against a backdrop of increased research focus on why certain emergent properties surface at scale, this work provides a useful counter-example. We posit that it is possible to optimize for a quantization friendly training recipe that suppresses large activation magnitude outliers. Here, we find that outlier dimensions are not an inherent product of scale, but rather sensitive to the optimization conditions present during pre-training. This both opens up directions for more efficient quantization, and poses the question of whether other emergent properties are inherent or can be altered and conditioned by optimization and architecture design choices. We successfully quantize models ranging in size from 410M to 52B with minimal degradation in performance.

NLP models often rely on superficial cues known as dataset biases to achieve impressive performance, and can fail on examples where these biases do not hold. Recent work sought to develop robust, unbiased models by filtering biased examples from training sets. In this work, we argue that such filtering can obscure the true capabilities of models to overcome biases, which might never be removed in full from the dataset. We suggest that in order to drive the development of models robust to subtle biases, dataset biases should be amplified in the training set. We introduce an evaluation framework defined by a bias-amplified training set and an anti-biased test set, both automatically extracted from existing datasets. Experiments across three notions of bias, four datasets and two models show that our framework is substantially more challenging for models than the original data splits, and even more challenging than hand-crafted challenge sets. Our evaluation framework can use any existing dataset, even those considered obsolete, to test model robustness. We hope our work will guide the development of robust models that do not rely on superficial biases and correlations. To this end, we publicly release our code and data.

We present a new distribution-free conformal prediction algorithm for sequential data (e.g., time series), called the \textit{sequential predictive conformal inference} (\texttt{SPCI}). We specifically account for the nature that time series data are non-exchangeable, and thus many existing conformal prediction algorithms are not applicable. The main idea is to adaptively re-estimate the conditional quantile of non-conformity scores (e.g., prediction residuals), upon exploiting the temporal dependence among them. More precisely, we cast the problem of conformal prediction interval as predicting the quantile of a future residual, given a user-specified point prediction algorithm. Theoretically, we establish asymptotic valid conditional coverage upon extending consistency analyses in quantile regression. Using simulation and real-data experiments, we demonstrate a significant reduction in interval width of \texttt{SPCI} compared to other existing methods under the desired empirical coverage.

GPT-3 and GPT-4 models are powerful, achieving high performance on a variety of Natural Language Processing tasks. However, there is a relative lack of detailed published analysis of their performance on the task of grammatical error correction (GEC). To address this, we perform experiments testing the capabilities of a GPT-3.5 model (text-davinci-003) and a GPT-4 model (gpt-4-0314) on major GEC benchmarks. We compare the performance of different prompts in both zero-shot and few-shot settings, analyzing intriguing or problematic outputs encountered with different prompt formats. We report the performance of our best prompt on the BEA-2019 and JFLEG datasets, finding that the GPT models can perform well in a sentence-level revision setting, with GPT-4 achieving a new high score on the JFLEG benchmark. Through human evaluation experiments, we compare the GPT models' corrections to source, human reference, and baseline GEC system sentences and observe differences in editing strategies and how they are scored by human raters.

As pointed out by several scholars, current research on hate speech (HS) recognition is characterized by unsystematic data creation strategies and diverging annotation schemata. Subsequently, supervised-learning models tend to generalize poorly to datasets they were not trained on, and the performance of the models trained on datasets labeled using different HS taxonomies cannot be compared. To ease this problem, we propose applying extremely weak supervision that only relies on the class name rather than on class samples from the annotated data. We demonstrate the effectiveness of a state-of-the-art weakly-supervised text classification model in various in-dataset and cross-dataset settings. Furthermore, we conduct an in-depth quantitative and qualitative analysis of the source of poor generalizability of HS classification models.

Large scale automatic software generation and configuration is a very complex and challenging task. In this proposal, we set out to investigate this problem by breaking down automatic software generation and configuration into two different tasks. In first task, we propose to synthesize software automatically with input output specifications. This task is further broken down into two sub-tasks. The first sub-task is about synthesizing programs with a genetic algorithm which is driven by a neural network based fitness function trained with program traces and specifications. For the second sub-task, we formulate program synthesis as a continuous optimization problem and synthesize programs with covariance matrix adaption evolutionary strategy (a state-of-the-art continuous optimization method). Finally, for the second task, we propose to synthesize configurations of large scale software from different input files (e.g.