In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities at many critical tasks. We argue that, without substantial effort to prevent it, AGIs could learn to pursue goals that are in conflict (i.e. misaligned) with human interests. If trained like today's most capable models, AGIs could learn to act deceptively to receive higher reward, learn misaligned internally-represented goals which generalize beyond their fine-tuning distributions, and pursue those goals using power-seeking strategies. We review emerging evidence for these properties. AGIs with these properties would be difficult to align and may appear aligned even when they are not. Finally, we briefly outline how the deployment of misaligned AGIs might irreversibly undermine human control over the world, and we review research directions aimed at preventing this outcome.
相關內容
State machines play a pivotal role in augmenting the efficacy of protocol analyzing to unveil more vulnerabilities. However, the task of inferring state machines from network protocol implementations presents significant challenges. Traditional methods based on dynamic analysis often overlook crucial state transitions due to limited coverage, while static analysis faces difficulties with complex code structures and behaviors. To address these limitations, we propose an innovative state machine inference approach powered by Large Language Models (LLMs). Utilizing text-embedding technology, this method allows LLMs to dissect and analyze the intricacies of protocol implementation code. Through targeted prompt engineering, we systematically identify and infer the underlying state machines. Our evaluation across six protocol implementations demonstrates the method's high efficacy, achieving an accuracy rate exceeding 90% and successfully delineating differences on state machines among various implementations of the same protocol. Importantly, integrating this approach with protocol fuzzing has notably enhanced AFLNet's code coverage by 10% over RFCNLP, showcasing the considerable potential of LLMs in advancing network protocol security analysis. Our proposed method not only marks a significant step forward in accurate state machine inference but also opens new avenues for improving the security and reliability of protocol implementations.
As artificially intelligent systems become more anthropomorphic and pervasive, and their potential impact on humanity more urgent, discussions about the possibility of machine consciousness have significantly intensified, and it is sometimes seen as 'the holy grail'. Many concerns have been voiced about the ramifications of creating an artificial conscious entity. This is compounded by a marked lack of consensus around what constitutes consciousness and by an absence of a universal set of criteria for determining consciousness. By going into depth on the foundations and characteristics of consciousness, we propose five criteria for determining whether a machine is conscious, which can also be applied more generally to any entity. This paper aims to serve as a primer and stepping stone for researchers of consciousness, be they in philosophy, computer science, medicine, or any other field, to further pursue this holy grail of philosophy, neuroscience and artificial intelligence.
In the past few decades, many multiobjective evolutionary optimization algorithms (MOEAs) have been proposed to find a finite set of approximate Pareto solutions for a given problem in a single run, each with its own structure. However, in many real-world applications, it could be desirable to have structure constraints on the entire optimal solution set, which define the patterns shared among all solutions. The current population-based MOEAs cannot properly handle such requirements. In this work, we make the first attempt to incorporate the structure constraints into the whole solution set by a single Pareto set model, which can be efficiently learned by a simple evolutionary stochastic optimization method. With our proposed method, the decision-makers can flexibly trade off the Pareto optimality with preferred structures among all solutions, which is not supported by previous MOEAs. A set of experiments on benchmark test suites and real-world application problems fully demonstrates the efficiency of our proposed method.
Since the emergence of GPT-3, Large Language Models (LLMs) have caught the eyes of researchers, practitioners, and educators in the field of software engineering. However, there has been relatively little investigation regarding the performance of LLMs in assisting with requirements analysis and UML modeling. This paper explores how LLMs can assist novice analysts in creating three types of typical UML models: use case models, class diagrams, and sequence diagrams. For this purpose, we designed the modeling tasks of these three UML models for 45 undergraduate students who participated in a requirements modeling course, with the help of LLMs. By analyzing their project reports, we found that LLMs can assist undergraduate students as notice analysts in UML modeling tasks, but LLMs also have shortcomings and limitations.
We introduce the task of human action anomaly detection (HAAD), which aims to identify anomalous motions in an unsupervised manner given only the pre-determined normal category of training action samples. Compared to prior human-related anomaly detection tasks which primarily focus on unusual events from videos, HAAD involves the learning of specific action labels to recognize semantically anomalous human behaviors. To address this task, we propose a normalizing flow (NF)-based detection framework where the sample likelihood is effectively leveraged to indicate anomalies. As action anomalies often occur in some specific body parts, in addition to the full-body action feature learning, we incorporate extra encoding streams into our framework for a finer modeling of body subsets. Our framework is thus multi-level to jointly discover global and local motion anomalies. Furthermore, to show awareness of the potentially jittery data during recording, we resort to discrete cosine transformation by converting the action samples from the temporal to the frequency domain to mitigate the issue of data instability. Extensive experimental results on two human action datasets demonstrate that our method outperforms the baselines formed by adapting state-of-the-art human activity AD approaches to our task of HAAD.
Recent artificial intelligence (AI) systems have reached milestones in "grand challenges" ranging from Go to protein-folding. The capability to retrieve medical knowledge, reason over it, and answer medical questions comparably to physicians has long been viewed as one such grand challenge. Large language models (LLMs) have catalyzed significant progress in medical question answering; Med-PaLM was the first model to exceed a "passing" score in US Medical Licensing Examination (USMLE) style questions with a score of 67.2% on the MedQA dataset. However, this and other prior work suggested significant room for improvement, especially when models' answers were compared to clinicians' answers. Here we present Med-PaLM 2, which bridges these gaps by leveraging a combination of base LLM improvements (PaLM 2), medical domain finetuning, and prompting strategies including a novel ensemble refinement approach. Med-PaLM 2 scored up to 86.5% on the MedQA dataset, improving upon Med-PaLM by over 19% and setting a new state-of-the-art. We also observed performance approaching or exceeding state-of-the-art across MedMCQA, PubMedQA, and MMLU clinical topics datasets. We performed detailed human evaluations on long-form questions along multiple axes relevant to clinical applications. In pairwise comparative ranking of 1066 consumer medical questions, physicians preferred Med-PaLM 2 answers to those produced by physicians on eight of nine axes pertaining to clinical utility (p < 0.001). We also observed significant improvements compared to Med-PaLM on every evaluation axis (p < 0.001) on newly introduced datasets of 240 long-form "adversarial" questions to probe LLM limitations. While further studies are necessary to validate the efficacy of these models in real-world settings, these results highlight rapid progress towards physician-level performance in medical question answering.
In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.
Emotion recognition in conversation (ERC) aims to detect the emotion label for each utterance. Motivated by recent studies which have proven that feeding training examples in a meaningful order rather than considering them randomly can boost the performance of models, we propose an ERC-oriented hybrid curriculum learning framework. Our framework consists of two curricula: (1) conversation-level curriculum (CC); and (2) utterance-level curriculum (UC). In CC, we construct a difficulty measurer based on "emotion shift" frequency within a conversation, then the conversations are scheduled in an "easy to hard" schema according to the difficulty score returned by the difficulty measurer. For UC, it is implemented from an emotion-similarity perspective, which progressively strengthens the model's ability in identifying the confusing emotions. With the proposed model-agnostic hybrid curriculum learning strategy, we observe significant performance boosts over a wide range of existing ERC models and we are able to achieve new state-of-the-art results on four public ERC datasets.
Deep neural networks (DNNs) are successful in many computer vision tasks. However, the most accurate DNNs require millions of parameters and operations, making them energy, computation and memory intensive. This impedes the deployment of large DNNs in low-power devices with limited compute resources. Recent research improves DNN models by reducing the memory requirement, energy consumption, and number of operations without significantly decreasing the accuracy. This paper surveys the progress of low-power deep learning and computer vision, specifically in regards to inference, and discusses the methods for compacting and accelerating DNN models. The techniques can be divided into four major categories: (1) parameter quantization and pruning, (2) compressed convolutional filters and matrix factorization, (3) network architecture search, and (4) knowledge distillation. We analyze the accuracy, advantages, disadvantages, and potential solutions to the problems with the techniques in each category. We also discuss new evaluation metrics as a guideline for future research.
Deep convolutional neural networks (CNNs) have recently achieved great success in many visual recognition tasks. However, existing deep neural network models are computationally expensive and memory intensive, hindering their deployment in devices with low memory resources or in applications with strict latency requirements. Therefore, a natural thought is to perform model compression and acceleration in deep networks without significantly decreasing the model performance. During the past few years, tremendous progress has been made in this area. In this paper, we survey the recent advanced techniques for compacting and accelerating CNNs model developed. These techniques are roughly categorized into four schemes: parameter pruning and sharing, low-rank factorization, transferred/compact convolutional filters, and knowledge distillation. Methods of parameter pruning and sharing will be described at the beginning, after that the other techniques will be introduced. For each scheme, we provide insightful analysis regarding the performance, related applications, advantages, and drawbacks etc. Then we will go through a few very recent additional successful methods, for example, dynamic capacity networks and stochastic depths networks. After that, we survey the evaluation matrix, the main datasets used for evaluating the model performance and recent benchmarking efforts. Finally, we conclude this paper, discuss remaining challenges and possible directions on this topic.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191