Ryan Steed, a PhD student at Carnegie Mellon University, and Aylin Caliskan, an assistant professor at George Washington University, looked at two algorithms: OpenAI's iGPT (a version of GPT-2 that is trained on pixels instead of words) and Google's SimCLR. While each algorithm approaches learning images differently, they share an important characteristic--they both use completely unsupervised learning, meaning they do not need humans to label the images. This is a relatively new innovation as of 2020. Previous computer-vision algorithms mainly used supervised learning, which involves feeding them manually labeled images: cat photos with the tag "cat" and baby photos with the tag "baby." But in 2019, researcher Kate Crawford and artist Trevor Paglen found that these human-created labels in ImageNet, the most foundational image data set for training computer-vision models, sometimes contain disturbing language, like "slut" for women and racial slurs for minorities.

After challenging the validity of these assumptions in real-world applications, we propose ways to move forward when they are violated. First, we show that group fairness criteria purely based on statistical properties of observed data are fundamentally limited. Revisiting this limitation from a causal viewpoint we develop a more versatile conceptual framework, causal fairness criteria, and first algorithms to achieve them. We also provide tools to analyze how sensitive a believed-to-be causally fair algorithm is to misspecifications of the causal graph. Second, we overcome the assumption that sensitive data is readily available in practice. To this end we devise protocols based on secure multi-party computation to train, validate, and contest fair decision algorithms without requiring users to disclose their sensitive data or decision makers to disclose their models. Finally, we also accommodate the fact that outcome labels are often only observed when a certain decision has been made. We suggest a paradigm shift away from training predictive models towards directly learning decisions to relax the traditional assumption that labels can always be recorded. The main contribution of this thesis is the development of theoretically substantiated and practically feasible methods to move research on fair machine learning closer to real-world applications.

We consider the problem of predicting the covariance of a zero mean Gaussian vector, based on another feature vector. We describe a covariance predictor that has the form of a generalized linear model, i.e., an affine function of the features followed by an inverse link function that maps vectors to symmetric positive definite matrices. The log-likelihood is a concave function of the predictor parameters, so fitting the predictor involves convex optimization. Such predictors can be combined with others, or recursively applied to improve performance.

Civil rights activists have successfully pushed for bans on police use of facial recognition in cities like Oakland, San Francisco, and Somerville, Massachusetts. Now, a coalition led by Amnesty International is setting its sights on the nation's biggest city--New York--as part of a drive for a global moratorium on government use of the technology. Amnesty's #BantheScan campaign is backed by Legal Aid, the New York Civil Liberties Union, and AI For the People among other groups. After New York, the group plans to target New Delhi and Ulaanbaatar in Mongolia. "New York is the biggest city in the country," says Michael Kleinman, director of Amnesty International's Silicon Valley Initiative.

Technology that measures emotions based on biometric indicators such as facial movements, tone of voice or body movements is increasingly being marketed in China, researchers say, despite concerns about its accuracy and wider human rights implications. Drawing upon artificial intelligence, the tools range from cameras to help police monitor a suspect's face during an interrogation to eye-tracking devices in schools that identify students who are not paying attention. A report released this week from U.K.-based human rights group Article 19 identified dozens of companies offering such tools in the education, public security and transportation sectors in China. "We believe that their design, development, deployment, sale and transfers should be banned due to the racist foundations and fundamental incompatibility with human rights," said Vidushi Marda, a senior program officer at Article 19. Human emotions cannot be reliably measured and quantified by technology tools, said Shazeda Ahmed, a doctoral candidate studying cybersecurity at the University of California, Berkeley and the report's co-author. Such systems can perpetuate bias, especially those sold to police that purport to identify criminality based on biometric indicators, she added.

Recent advances in deep learning techniques have enabled machines to generate cohesive open-ended text when prompted with a sequence of words as context. While these models now empower many downstream applications from conversation bots to automatic storytelling, they have been shown to generate texts that exhibit social biases. To systematically study and benchmark social biases in open-ended language generation, we introduce the Bias in Open-Ended Language Generation Dataset (BOLD), a large-scale dataset that consists of 23,679 English text generation prompts for bias benchmarking across five domains: profession, gender, race, religion, and political ideology. We also propose new automated metrics for toxicity, psycholinguistic norms, and text gender polarity to measure social biases in open-ended text generation from multiple angles. An examination of text generated from three popular language models reveals that the majority of these models exhibit a larger social bias than human-written Wikipedia text across all domains. With these results we highlight the need to benchmark biases in open-ended language generation and caution users of language generation models on downstream tasks to be cognizant of these embedded prejudices.

Claim: fin(tech) offers lessons for many areas in law 163.

Stronger action needs to be taken to stop technologies like facial recognition from being used to violate fundamental human rights, because the ethics charters currently adopted by businesses and governments won't cut it, warns a new report from digital rights organization Access Now. The past few years have seen "ethical AI" become a hot topic, with requirements such as oversight, safety, privacy, transparency, or accountability being added to codes of conduct for private and public organizations alike. From 5% in 2019, in fact, the proportion of organizations that now have an AI ethics charter has jumped to 45% in 2020. The EU's guidelines for "Trustworthy AI" have informed many of these documents; in addition, the European bloc recently published a white paper on artificial intelligence presenting a so-called "European framework for AI", with ethics at its core. How much real change has happened as a result of those ethical guidelines is up for debate.

Facial recognition technology amplifies racist policing, threatens the right to protest and should be banned globally, Amnesty International said as it urged New York City to pass a ban on its use in mass surveillance by law enforcement. "Facial recognition risks being weaponised by law enforcement against marginalised communities around the world," said Matt Mahmoudi, AI and human rights researcher at Amnesty. "From New Delhi to New York, this invasive technology turns our identities against us and undermines human rights. "New Yorkers should be able to go out about their daily lives without being tracked by facial recognition. Other major cities across the US have already banned facial recognition, and New York must do the same." Albert Fox Cahn of New York's Urban Justice Centre, which is supporting Amnesty's Ban the Scan campaign, said: "Facial recognition is biased, broken, and antithetical to democracy.

Facial recognition technology amplifies racist policing, threatens the right to protest and should be banned globally, Amnesty International said as it urged New York City to pass a ban on its use in mass surveillance by law enforcement. "Facial recognition risks being weaponised by law enforcement against marginalised communities around the world," said Matt Mahmoudi, AI and human rights researcher at Amnesty. "From New Delhi to New York, this invasive technology turns our identities against us and undermines human rights. "New Yorkers should be able to go out about their daily lives without being tracked by facial recognition. Other major cities across the US have already banned facial recognition, and New York must do the same." Albert Fox Cahn of New York's Urban Justice Centre, which is supporting Amnesty's Ban the Scan campaign, said: "Facial recognition is biased, broken, and antithetical to democracy.