The soap opera-like ousting and swift return of OpenAI CEO Sam Altman produced plenty of fodder for ironic quips online but it also exposed some serious fault lines. One such critique I enjoyed was: "How are we supposed to solve the AI alignment problem if aligning just a few board members presents an insurmountable challenge?" As the company behind ChatGPT, OpenAI may be one of the more recognizable names, but artificial intelligence is more than one company. It's a technology of immense consequence, yet it remains almost entirely unregulated. The E.U. has a chance to meaningfully tackle that challenge--but not if it bends the knee to Big Tech's ongoing onslaught. Inspirational Members of the European Parliament have so far been standing firm in the face of incredible pressure, in an effort to save this landmark legislation.

This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization.

In economic modeling, there has been an increasing investigation into multi-agent simulators. Nevertheless, state-of-the-art studies establish the model based on reinforcement learning (RL) exclusively for specific agent categories, e.g., households, firms, or the government. It lacks concerns over the resulting adaptation of other pivotal agents, thereby disregarding the complex interactions within a real-world economic system. Furthermore, we pay attention to the vital role of the government policy in distributing tax credits. Instead of uniform distribution considered in state-of-the-art, it requires a well-designed strategy to reduce disparities among households and improve social welfare. To address these limitations, we propose an expansive multi-agent economic model comprising reinforcement learning agents of numerous types. Additionally, our research comprehensively explores the impact of tax credit allocation on household behavior and captures the spectrum of spending patterns that can be observed across diverse households. Further, we propose an innovative government policy to distribute tax credits, strategically leveraging insights from tax credit spending patterns. Simulation results illustrate the efficacy of the proposed government strategy in ameliorating inequalities across households.

The past decade has observed a great advancement in AI with deep learning-based models being deployed in diverse scenarios including safety-critical applications. As these AI systems become deeply embedded in our societal infrastructure, the repercussions of their decisions and actions have significant consequences, making the ethical implications of AI deployment highly relevant and important. The ethical concerns associated with AI are multifaceted, including challenging issues of fairness, privacy and data protection, responsibility and accountability, safety and robustness, transparency and explainability, and environmental impact. These principles together form the foundations of ethical AI considerations that concern every stakeholder in the AI system lifecycle. In light of the present ethical and future x-risk concerns, governments have shown increasing interest in establishing guidelines for the ethical deployment of AI. This work unifies the current and future ethical concerns of deploying AI into society. While we acknowledge and appreciate the technical surveys for each of the ethical principles concerned, in this paper, we aim to provide a comprehensive overview that not only addresses each principle from a technical point of view but also discusses them from a social perspective.

NLP models are used in a variety of critical social computing tasks, such as detecting sexist, racist, or otherwise hateful content. Therefore, it is imperative that these models are robust to spurious features. Past work has attempted to tackle such spurious features using training data augmentation, including Counterfactually Augmented Data (CADs). CADs introduce minimal changes to existing training data points and flip their labels; training on them may reduce model dependency on spurious features. However, manually generating CADs can be time-consuming and expensive. Hence in this work, we assess if this task can be automated using generative NLP models. We automatically generate CADs using Polyjuice, ChatGPT, and Flan-T5, and evaluate their usefulness in improving model robustness compared to manually-generated CADs. By testing both model performance on multiple out-of-domain test sets and individual data point efficacy, our results show that while manual CADs are still the most effective, CADs generated by ChatGPT come a close second. One key reason for the lower performance of automated methods is that the changes they introduce are often insufficient to flip the original label.

Given the growing importance of AI literacy, we decided to write this tutorial to help narrow the gap between the discourse among those who study language models -- the core technology underlying ChatGPT and similar products -- and those who are intrigued and want to learn more about them. In short, we believe the perspective of researchers and educators can add some clarity to the public's understanding of the technologies beyond what's currently available, which tends to be either extremely technical or promotional material generated about products by their purveyors. Our approach teases apart the concept of a language model from products built on them, from the behaviors attributed to or desired from those products, and from claims about similarity to human cognition. As a starting point, we (1) offer a scientific viewpoint that focuses on questions amenable to study through experimentation; (2) situate language models as they are today in the context of the research that led to their development; and (3) describe the boundaries of what is known about the models at this writing.

Automated machine learning (AutoML) gathered a significant amount of attention in recent years as a way of automating some of the typical workflows in machine learning (ML) and data science more broadly. For a comprehensive and systematic view on the subject, there is an already growing number of survey works that cover the state-of-the-art Hutter et al. (2019); Yao et al. (2018); Elshawi et al. (2019); Zöller and Huber (2021); Truong et al. (2019); He et al. (2021); Hospedales et al. (2020); Vanschoren (2018Santu"); Karmaker Santu"Santu". Currently, it is becoming apparent that the size of the potential problem space, required solution sophistication, transparency and legal constraints Roscher et al. (2020); Drozdal et al. (2020); Rudin et al. (2021); Veale and Borgesius (2021); Smuha et al. (2021) render AutoML a problem extremely difficult to define and solve either holistically or agnostically.

Solving the AI alignment problem requires having clear, defensible values towards which AI systems can align. Currently, targets for alignment remain underspecified and do not seem to be built from a philosophically robust structure. We begin the discussion of this problem by presenting five core, foundational values, drawn from moral philosophy and built on the requisites for human existence: survival, sustainable intergenerational existence, society, education, and truth. We show that these values not only provide a clearer direction for technical alignment work, but also serve as a framework to highlight threats and opportunities from AI systems to both obtain and sustain these values.

Machine translation (MT) is routinely evaluated using various segment-level similarity metrics against one or more reference translations. At the same time, reference translations acquired in the standard way are often criticized for their flaws of various types. For several high-resourced language pairs, MT quality reaches levels comparable to the quality of the reference translation (Freitag et al. 2022; Hassan et al. 2018) and sometimes MT even significantly surpasses humans in a particular evaluation setting (Popel et al. 2020). Given this, one could conclude that state-of-the-art MT has reached the point where reference-based evaluation is no longer reliable and we have to resort to other methods (such as targeted expert evaluation of particular outputs), even if they are costly, subjective and possibly impossible to automate. The narrow goal of the presented work is to allow for an "extension of the expiry date" for reference-based evaluation methods. In a broader perspective, we want to formulate a methodology for creating reference translations which avoid the often-observed deficiencies of "standard" or "professional" reference translations, be it multiple interfering phenomena, inappropriate expressions, ignorance of topic-focus articulation (information structure) or other abundant shortcomings in the translation, indicating their authors' insensitivity to the topic itself, but above all to the source and target language. To this end, we introduce so-called optimal reference translations (ORT), which are intended to represent optimal (ideal or excellent) human translations (should they be the subject of a translation quality evaluation).

Algorithmic fairness is of utmost societal importance, yet the current trend in large-scale machine learning models requires training with massive datasets that are frequently biased. In this context, pre-processing methods that focus on modeling and correcting bias in the data emerge as valuable approaches. In this paper, we propose FairShap, a novel instance-level data re-weighting method for fair algorithmic decision-making through data valuation by means of Shapley Values. FairShap is model-agnostic and easily interpretable, as it measures the contribution of each training data point to a predefined fairness metric. We empirically validate FairShap on several state-of-the-art datasets of different nature, with a variety of training scenarios and models and show how it yields fairer models with similar levels of accuracy than the baselines. We illustrate FairShap's interpretability by means of histograms and latent space visualizations. Moreover, we perform a utility-fairness study, and ablation and runtime experiments to illustrate the impact of the size of the reference dataset and FairShap's computational cost depending on the size of the dataset and the number of features. We believe that FairShap represents a promising direction in interpretable and model-agnostic approaches to algorithmic fairness that yield competitive accuracy even when only biased datasets are available.