For those engaged in advocacy around the social harms of AI systems, a definitional exercise could, however, be a key way to rescue AI from the abstract, and foreground social and material concerns around these systems. Just as glossy data visualizations can obscure the unequal impacts and governance failures of the pandemic, AI as an abstract buzzword can be brandished against complex social problems as if it were a neutral and external'solution' rather than a sociotechnical system 14 designed and developed to make value-laden choices and trade-offs.

Europe's vision of artificial intelligence regulation is color-blind -- and not in a good way. Between the U.S.'s laissez-faire and China's dirigiste approaches, the EU is intent on carving out a "third way" for AI regulation that boosts innovation but respects "European values," including privacy and human rights. But activists and academics fear the rules will not consider the communities most at risk of AI-based discrimination -- people of color. In recent years, there have been high-profile examples of AI systems discriminating against racial minorities, including facial recognition systems that don't recognize women or black and brown faces; opaque, unenforceable and discriminatory hiring algorithms; or applications that predict disproportionate criminality and offer worse legal outcomes. The European Commission will unveil its AI rules this spring, requiring "high-risk" AI systems to meet minimum standards regarding trustworthiness.

Machine Learning (ML) seems to be one of the most promising solution to automate partially or completely some of the complex tasks currently realized by humans, such as driving vehicles, recognizing voice, etc. It is also an opportunity to implement and embed new capabilities out of the reach of classical implementation techniques. However, ML techniques introduce new potential risks. Therefore, they have only been applied in systems where their benefits are considered worth the increase of risk. In practice, ML techniques raise multiple challenges that could prevent their use in systems submitted to certification constraints. But what are the actual challenges? Can they be overcome by selecting appropriate ML techniques, or by adopting new engineering or certification practices? These are some of the questions addressed by the ML Certification 3 Workgroup (WG) set-up by the Institut de Recherche Technologique Saint Exup\'ery de Toulouse (IRT), as part of the DEEL Project.

Now that science fiction is reality, it's time for new laws of robotics Eighty years ago, Isaac Asimov dreamed up three rules to ensure machines would serve humanity. It's a framework that has shaped decades of debates about AI ethics, but it needs an update. Here are four laws of robotics for the real world.

Developing and deploying AI-powered systems and applications is a complex business, especially in our extended remote reality. You're likely facing an uphill climb and let's face it, huge risks. The way to clear the obstacles, lower the risks, and raise the value you deliver hinges on one essential element: implementing DevSecOps to protect your process and your assets. We're operating in a different world now where unity among development (Dev), security (Sec), and operations (Ops) has never been more essential. Compounded by pressure related to the fast need to convert many of our office infrastructure to meet the needs of our remote reality during the COVID-19 pandemic, the market for DevSecOps is projected to grow from 32% to 34% mid-decade.[i]

AI, or artificial intelligence, seems to be on the tip of everyone's tongue these days. While I've been aware of this major trend in tech development for a while, I've noticed AI appearing more and more as one of the most in-demand areas of expertise for job seekers. I'm sure that for many of us, the term "AI" conjures up sci-fi fantasies or fear about robots taking over the world. The depictions of AI in the media have run the gamut, and while no one can predict exactly how it will evolve in the future, the current trends and developments paint a much different picture of how AI will become part of our lives. In reality, AI is already at work all around us, impacting everything from our search results, to our online dating prospects, to the way we shop.

My work at OpenAI reminds me every day about the magnitude of the socioeconomic change that is coming sooner than most people believe. Software that can think and learn will do more and more of the work that people now do. Even more power will shift from labor to capital. If public policy doesn't adapt accordingly, most people will end up worse off than they are today. We need to design a system that embraces this technological future and taxes the assets that will make up most of the value in that world–companies and land–in order to fairly distribute some of the coming wealth.

Real-world decision-making tasks are generally complex, requiring trade-offs between multiple, often conflicting, objectives. Despite this, the majority of research in reinforcement learning and decision-theoretic planning either assumes only a single objective, or that multiple objectives can be adequately handled via a simple linear combination. Such approaches may oversimplify the underlying problem and hence produce suboptimal results. This paper serves as a guide to the application of multi-objective methods to difficult problems, and is aimed at researchers who are already familiar with single-objective reinforcement learning and planning methods who wish to adopt a multi-objective perspective on their research, as well as practitioners who encounter multi-objective decision problems in practice. It identifies the factors that may influence the nature of the desired solution, and illustrates by example how these influence the design of multi-objective decision-making systems for complex problems.

Existing survival analysis techniques heavily rely on strong modelling assumptions and are, therefore, prone to model misspecification errors. In this paper, we develop an inferential method based on ideas from conformal prediction, which can wrap around any survival prediction algorithm to produce calibrated, covariate-dependent lower predictive bounds on survival times. In the Type I right-censoring setting, when the censoring times are completely exogenous, the lower predictive bounds have guaranteed coverage in finite samples without any assumptions other than that of operating on independent and identically distributed data points. Under a more general conditionally independent censoring assumption, the bounds satisfy a doubly robust property which states the following: marginal coverage is approximately guaranteed if either the censoring mechanism or the conditional survival function is estimated well. Further, we demonstrate that the lower predictive bounds remain valid and informative for other types of censoring. The validity and efficiency of our procedure are demonstrated on synthetic data and real COVID-19 data from the UK Biobank.

In modern litigation, fraud investigators often face an overwhelming number of documents that must be reviewed throughout a matter. In the majority of legal cases, fraud investigators do not know beforehand, exactly what they are looking for, nor where to find it. In addition, fraudsters may use deception to hide their behaviour and intentions by using code words. Effectively, this means fraud investigators are looking for a needle in the haystack without knowing what the needle looks like. As part of a larger research program, we use a framework to expedite the investigation process applying text-mining and machine learning techniques. We structure this framework using three well-known methods in fraud investigations: (i) the fraud triangle (ii) the golden ("W") investigation questions, and (iii) the analysis of competing hypotheses. With this framework, it is possible to automatically organize investigative data, so it is easier for investigators to find answers to typical investigative questions. In this research, we focus on one of the components of this framework: the identification of the usage of code words by fraudsters. Here for, a novel (annotated) synthetic data set is created containing such code words, hidden in normal email communication. Subsequently, a range of machine learning techniques are employed to detect such code words. We show that the state-of-the-art BERT model significantly outperforms other methods on this task. With this result, we demonstrate that deep neural language models can reliably (F1 score of 0.9) be applied in fraud investigations for the detection of code words.