Where will we draw the line between human-generated art and AI-generated art? If so, should we bestow the acclaimed title of artisan upon said AI? Let's unpack things and see where the world stands on these mind-bending concerns. A crucial undercurrent has to do with AI Ethics and how we as a society perceive and want to make use of AI. For my ongoing and extensive coverage of AI Ethics and Ethical AI, see the link here and the link here, just to name a few. News stories this past few days have made AI and art an extremely hot topic. You see, the whole conundrum about Artificial Intelligence and art was recently thrust into the public eye when an AI "artbot" seemingly won an art contest. The headlines regarding this matter have ranged from fervent outrage to a sense of sorrowful acquiescence that it was only a matter of time before AI would prevail in the creative field of artistry. Some even claim that we've already seen AI comeuppance in art and that there is nothing new in this latest occurrence other than it managed to touch a nerve on social media. Amid all the heated debate in general, there are a lot of facts about this latest incident that muddy the waters and tend to undercut the shallow headlines and vitriolic tweets that the story has generated. It might be useful to take a moment and calmly consider the actual specifics, which I will be doing throughout this discussion. Meanwhile, one perhaps beneficial outcome of the reported story is that AI Ethics managed to suddenly get some long overdue recognition in the media at large. Whenever an AI-themed man-bites-dog story hits the airwaves and goes viral on social media, public opinions start to weigh in. We will examine the various qualms and complaints expressed in the public discourse about this brewing AI Ethics riddle. First, let's lay out the facts of the deemed newsworthy snowball that ultimately started a cantankerous snowfall avalanche. The Colorado State Fair is where the competition in this case took place. The Fair is an annual event that has a hearty 150-year-old tradition initially focused on livestock. An eventual expansion of activities included the inclusion of a fine arts contest.

Federated learning (FL) has emerged as a practical solution to tackle data silo issues without compromising user privacy. One of its variants, vertical federated learning (VFL), has recently gained increasing attention as the VFL matches the enterprises' demands of leveraging more valuable features to build better machine learning models while preserving user privacy. Current works in VFL concentrate on developing a specific protection or attack mechanism for a particular VFL algorithm. In this work, we propose an evaluation framework that formulates the privacy-utility evaluation problem. We then use this framework as a guide to comprehensively evaluate a broad range of protection mechanisms against most of the state-of-the-art privacy attacks for three widely-deployed VFL algorithms. These evaluations may help FL practitioners select appropriate protection mechanisms given specific requirements. Our evaluation results demonstrate that: the model inversion and most of the label inference attacks can be thwarted by existing protection mechanisms; the model completion (MC) attack is difficult to be prevented, which calls for more advanced MC-targeted protection mechanisms. Based on our evaluation results, we offer concrete advice on improving the privacy-preserving capability of VFL systems.

ABSTRACT Human trafficking is a universal problem, persistent despite numerous efforts to combat globally. Individuals of any age, race, ethnicity, sex, gender identity, sexual orientation, nationality, immigration status, cultural background, religion, socio-economic class, and education can be a victim of human trafficking. With the advancements in technology and the introduction of autonomous vehicles (AVs), human traffickers will adopt new ways to transport victims, which could accelerate the growth of organized human trafficking networks, whcih can make detection of trafficking in persons more challenging for law enforcement agencies. The objective of this study is to develop an innovative audio analytics-based human trafficking detection framework for autonomous vehicles. The primary contributions of this study are to: (i) define four non-trivial, feasible, and realistic human trafficking scenarios for AVs; (ii) create a new and comprehensive audio dataset related to human trafficking with five classes--i.e., crying, screaming, car door banging, car noise, and conversation; and (iii) develop a deep 1-D Convolution Neural Network (CNN) architecture for audio data classification related to human trafficking. We have also conducted a case study using the new audio dataset and evaluate the audio classification performance of the deep 1-D CNN. Our analyses reveal that the deep 1-D CNN can distinguish sound coming from a human trafficking victim from a non-human trafficking sound with an accuracy of 95%, which proves the efficacy of our framework. INTRODUCTION Human trafficking is a global epidemic. People of any age, gender identities, and ethnicities from all across the world are constantly under threat of being victim of human trafficking. According to the Department of Homeland Security, falsification or threat of force is used to acquire cheap labor or commercial sex acts in human trafficking (1).

Datasets scraped from the internet have been critical to the successes of large-scale machine learning. Yet, this very success puts the utility of future internet-derived datasets at potential risk, as model outputs begin to replace human annotations as a source of supervision. In this work, we first formalize a system where interactions with one model are recorded as history and scraped as training data in the future. We then analyze its stability over time by tracking changes to a test-time bias statistic (e.g. gender bias of model predictions). We find that the degree of bias amplification is closely linked to whether the model's outputs behave like samples from the training distribution, a behavior which we characterize and define as consistent calibration. Experiments in three conditional prediction scenarios - image classification, visual role-labeling, and language generation - demonstrate that models that exhibit a sampling-like behavior are more calibrated and thus more stable. Based on this insight, we propose an intervention to help calibrate and stabilize unstable feedback systems. Code is available at https://github.com/rtaori/data_feedback.

In this meta-ethnography, we explore three different angles of ethical artificial intelligence (AI) design implementation including the philosophical ethical viewpoint, the technical perspective, and framing through a political lens. Our qualitative research includes a literature review which highlights the cross referencing of these angles through discussing the value and drawbacks of contrastive top-down, bottom-up, and hybrid approaches previously published. The novel contribution to this framework is the political angle, which constitutes ethics in AI either being determined by corporations and governments and imposed through policies or law (coming from the top), or ethics being called for by the people (coming from the bottom), as well as top-down, bottom-up, and hybrid technicalities of how AI is developed within a moral construct and in consideration of its users, with expected and unexpected consequences and long-term impact in the world. There is a focus on reinforcement learning as an example of a bottom-up applied technical approach and AI ethics principles as a practical top-down approach. This investigation includes real-world case studies to impart a global perspective, as well as philosophical debate on the ethics of AI and theoretical future thought experimentation based on historical fact, current world circumstances, and possible ensuing realities.

The government is facing a highly unusual five-day legal challenge to the policy involving at least 10 migrants, campaign groups Care4Calais and Detention Action, and the Public and Commercial Services Union, which represents the vast majority of UK Border Force staff.

We've found we can improve language model behavior with respect to specific behavioral values by fine-tuning on a curated dataset of 100 examples of those values. We also found that this process becomes more effective as models get larger. While the technique is still nascent, we're looking for OpenAI API users who would like to try it out and are excited to find ways to use these and other techniques in production use cases. Our approach aims to give language model operators the tools to narrow this universal set of behaviors to a constrained set of values. While OpenAI provides guardrails and monitoring to ensure that model use-cases are compatible with our Charter, we view selecting the exact set of Charter-compatible values for the model as a choice that our users must face for their specific applications.

That can be a problem when trying to train models that might benefit from more data, but regulatory issues restrict that data's movements, according to Steve Irvine, co-founder and CEO of integrate.ai. "[For] a lot of industries, like health care, it's prohibited moving the data across jurisdiction, and so some of the most meaningful use cases that you and I would hope could come into the world -- and developers want to bring into the world -- are blocked because the data can't move," Irvine said. This is where federated learning can help. Federated learning allows for the training of AI models by shifting the paradigm to bring the training function to the data, Irvine told The New Stack. "Instead of data having to come to a central location to train the machine learning model, versions of the model gets sent out to the location where the data resides," he explained.

How to integrate an emerging and all-pervasive technology such as AI into the structures and operations of our society is a question of contemporary politics, science and public debate. It has produced a considerable amount of international academic literature from different disciplines. This article analyzes the academic debate around the regulation of artificial intelligence (AI). The systematic review comprises a sample of 73 peer-reviewed journal articles published between January 1st, 2016, and December 31st, 2020. The analysis concentrates on societal risks and harms, questions of regulatory responsibility, and possible adequate policy frameworks, including risk-based and principle-based approaches. The main interests are proposed regulatory approaches and instruments. Various forms of interventions such as bans, approvals, standard-setting, and disclosure are presented. The assessments of the included papers indicate the complexity of the field, which shows its prematurity and the remaining lack of clarity. By presenting a structured analysis of the academic debate, we contribute both empirically and conceptually to a better understanding of the nexus of AI and regulation and the underlying normative decisions. A comparison of the scientific proposals with the proposed European AI regulation illustrates the specific approach of the regulation, its strengths and weaknesses.

Online trolls increase social costs and cause psychological damage to individuals. With the proliferation of automated accounts making use of bots for trolling, it is difficult for targeted individual users to handle the situation both quantitatively and qualitatively. To address this issue, we focus on automating the method to counter trolls, as counter responses to combat trolls encourage community users to maintain ongoing discussion without compromising freedom of expression. For this purpose, we propose a novel dataset for automatic counter response generation. In particular, we constructed a pair-wise dataset that includes troll comments and counter responses with labeled response strategies, which enables models fine-tuned on our dataset to generate responses by varying counter responses according to the specified strategy. We conducted three tasks to assess the effectiveness of our dataset and evaluated the results through both automatic and human evaluation. In human evaluation, we demonstrate that the model fine-tuned on our dataset shows a significantly improved performance in strategy-controlled sentence generation.