Large Language Models (LLMs) are foundational to AI advancements, facilitating applications like predictive text generation. Nonetheless, they pose risks by potentially memorizing and disseminating sensitive, biased, or copyrighted information from their vast datasets. Machine unlearning emerges as a cutting-edge solution to mitigate these concerns, offering techniques for LLMs to selectively discard certain data. This paper reviews the latest in machine unlearning for LLMs, introducing methods for the targeted forgetting of information to address privacy, ethical, and legal challenges without necessitating full model retraining. It divides existing research into unlearning from unstructured/textual data and structured/classification data, showcasing the effectiveness of these approaches in removing specific data while maintaining model efficacy. Highlighting the practicality of machine unlearning, this analysis also points out the hurdles in preserving model integrity, avoiding excessive or insufficient data removal, and ensuring consistent outputs, underlining the role of machine unlearning in advancing responsible, ethical AI.

Large Language Models (LLMs) exhibit impressive capabilities but also present risks such as biased content generation and privacy issues. One of the current alignment techniques includes principle-driven integration, but it faces challenges arising from the imprecision of manually crafted rules and inadequate risk perception in models without safety training. To address these, we introduce Guide-Align, a two-stage approach. Initially, a safety-trained model identifies potential risks and formulates specific guidelines for various inputs, establishing a comprehensive library of guidelines and a model for input-guidelines retrieval. Subsequently, the retrieval model correlates new inputs with relevant guidelines, which guide LLMs in response generation to ensure safe and high-quality outputs, thereby aligning with human values. An additional optional stage involves fine-tuning a model with well-aligned datasets generated through the process implemented in the second stage. Our method customizes guidelines to accommodate diverse inputs, thereby enhancing the fine-grainedness and comprehensiveness of the guideline library. Furthermore, it incorporates safety expertise from a safety-trained LLM through a lightweight retrieval model. We evaluate our approach on three benchmarks, demonstrating significant improvements in LLM security and quality. Notably, our fine-tuned model, Labrador, even at 13 billion parameters, outperforms GPT-3.5-turbo and surpasses GPT-4 in alignment capabilities.

Heritage Foundation tech policy director Kara Frederick joins'America's Newsroom' to discuss pornographic AI photos of Taylor Swift sparking conversations about deepfake regulation. Italian Premier Giorgia Meloni has been asked to testify in court July 2 in the trial of two men who are accused of making deepfake pornographic images using her face and posting them online. Meloni, who is listed as an injured party in the trial in Sassari in Sardinia, is seeking 108,212 in symbolic damages and will donate any award to an Interior Ministry fund for women victims of domestic violence, her attorney Maria Giulia Marongiu said in an email Friday to The Associated Press. "The crime in question is particularly odious, as it allegedly involves the uploading of fabricated pornographic images that could affect any unsuspecting woman with damaging consequences for her reputation and private life," Marongiu said. Italian Premier Giorgia Meloni arrives in the courtyard of the Italian government office Chigi Palace, to meet Kazakhstan President Kassym-Jomart Tokayev, in Rome, on Jan. 18, 2024. Meloni has been asked to testify in court on July 2, 2024, in the trial of two men who are accused of making deepfake pornographic images using her face and posting them online.

The Department of Justice and more than a dozen states have filed a lawsuit against Apple in the US federal court, accusing the company of violating antitrust laws. It says Apple's hardware and software products are largely inaccessible to competitors, making it difficult for rivals to compete and for customers to switch to other companies' products. The lawsuit comes after the European Commission fined Apple 1.8 billion ( 1.95 billion) for stopping music-streaming developers from "informing iOS users about alternative and cheaper music subscription services available" outside the App Store. The DOJ suggests Apple used its control over iOS to block innovative apps and cloud streaming services from the public. The suit also suggests Apple has obstructed rival payment platforms, made it harder for Android messages to appear on iPhones and restricted how competing smartphones integrated with iOS devices.

The advent of big data, and of Artificial Intelligence (AI) applications that collect and consume such data, has led to fundamental questions about the ethics of AI designs and to efforts aimed to highlight and safeguard against any potential harms caused by the deployment of AI across diverse domains of applications. Typically, questions raised relate to the trustworthiness of AI as agent technologies that autonomously or semi-autonomously operate in human environments and that have the ability to alter human behaviour. Other questions concern the role that AI may play now and in the future in either resolving or amplifying pre-existing social biases and any resulting harms. Specifically, Ethical AI as an emergent area of AI research and policy, has been spurred by the revelations of AI applications (usually unintentionally) promoting and amplifying many of the discriminatory and oppressive practices, and assumptions that underpin pre-existing social and institutional systems, e.g., historical biases against non-dominant populations, against users characterised by some divergence from the so-called cognitive or physical'norm', or those who are socio-economically disadvantaged (Crawford, 2017a; Madaio et al., 2022; Porayska-Pomsta and Rajendran, 2019; Williamson, Eynon, Knox & Davis, in this volume). Numerous examples of AI bias are both well-documented and rehearsed throughout the emergent ethics of AI literature, in hundreds of policy reports about AI ethics and governance that have been published to date (c.f.

This paper examines the role that enchantment plays in the design of AI things by constructing a taxonomy of design approaches that increase or decrease the perception of magic and enchantment. We start from the design discourse surrounding recent developments in AI technologies, highlighting specific interaction qualities such as algorithmic uncertainties and errors and articulating relations to the rhetoric of magic and supernatural thinking. Through analyzing and reflecting upon 52 students' design projects from two editions of a Master course in design and AI, we identify seven design principles and unpack the effects of each in terms of enchantment and disenchantment. We conclude by articulating ways in which this taxonomy can be approached and appropriated by design/HCI practitioners, especially to support exploration and reflexivity.

Domestic violence, which is often perceived as a gendered issue among female victims, has gained increasing attention in recent years. Despite this focus, male victims of domestic abuse remain primarily overlooked, particularly in Bangladesh. Our study represents a pioneering exploration of the underexplored realm of male domestic violence (MDV) within the Bangladeshi context, shedding light on its prevalence, patterns, and underlying factors. Existing literature predominantly emphasizes female victimization in domestic violence scenarios, leading to an absence of research on male victims. We collected data from the major cities of Bangladesh and conducted exploratory data analysis to understand the underlying dynamics. We implemented 11 traditional machine learning models with default and optimized hyperparameters, 2 deep learning, and 4 ensemble models. Despite various approaches, CatBoost has emerged as the top performer due to its native support for categorical features, efficient handling of missing values, and robust regularization techniques, achieving 76% accuracy. In contrast, other models achieved accuracy rates in the range of 58-75%. The eXplainable AI techniques, SHAP and LIME, were employed to gain insights into the decision-making of black-box machine learning models. By shedding light on this topic and identifying factors associated with domestic abuse, the study contributes to identifying groups of people vulnerable to MDV, raising awareness, and informing policies and interventions aimed at reducing MDV. Our findings challenge the prevailing notion that domestic abuse primarily affects women, thus emphasizing the need for tailored interventions and support systems for male victims. ML techniques enhance the analysis and understanding of the data, providing valuable insights for developing effective strategies to combat this pressing social issue.

Pretrained large language models (LLMs) are currently state-of-the-art for solving the vast majority of natural language processing tasks. While many real-world applications still require fine-tuning to reach satisfactory levels of performance, many of them are in the low-data regime, making fine-tuning challenging. To address this, we propose LLM2LLM, a targeted and iterative data augmentation strategy that uses a teacher LLM to enhance a small seed dataset by augmenting additional data that can be used for fine-tuning on a specific task. LLM2LLM (1) fine-tunes a baseline student LLM on the initial seed data, (2) evaluates and extracts data points that the model gets wrong, and (3) uses a teacher LLM to generate synthetic data based on these incorrect data points, which are then added back into the training data. This approach amplifies the signal from incorrectly predicted data points by the LLM during training and reintegrates them into the dataset to focus on more challenging examples for the LLM. Our results show that LLM2LLM significantly enhances the performance of LLMs in the low-data regime, outperforming both traditional fine-tuning and other data augmentation baselines. LLM2LLM reduces the dependence on labor-intensive data curation and paves the way for more scalable and performant LLM solutions, allowing us to tackle data-constrained domains and tasks. We achieve improvements up to 24.2% on the GSM8K dataset, 32.6% on CaseHOLD, 32.0% on SNIPS, 52.6% on TREC and 39.8% on SST-2 over regular fine-tuning in the low-data regime using a LLaMA2-7B student model.

This paper explores the pressing issue of risk assessment in Large Language Models (LLMs) as they become increasingly prevalent in various applications. Focusing on how reward models, which are designed to fine-tune pretrained LLMs to align with human values, perceive and categorize different types of risks, we delve into the challenges posed by the subjective nature of preference-based training data. By utilizing the Anthropic Red-team dataset, we analyze major risk categories, including Information Hazards, Malicious Uses, and Discrimination/Hateful content. Our findings indicate that LLMs tend to consider Information Hazards less harmful, a finding confirmed by a specially developed regression model. Additionally, our analysis shows that LLMs respond less stringently to Information Hazards compared to other risks. The study further reveals a significant vulnerability of LLMs to jailbreaking attacks in Information Hazard scenarios, highlighting a critical security concern in LLM risk assessment and emphasizing the need for improved AI safety measures.

With the recent development of large language models (LLMs), models that focus on certain domains and languages have been discussed for their necessity. There is also a growing need for benchmarks to evaluate the performance of current LLMs in each domain. Therefore, in this study, we constructed a benchmark comprising multiple tasks specific to the Japanese and financial domains and performed benchmark measurements on some models. Consequently, we confirmed that GPT-4 is currently outstanding, and that the constructed benchmarks function effectively. According to our analysis, our benchmark can differentiate benchmark scores among models in all performance ranges by combining tasks with different difficulties.