Alphabet, the parent company of Google GOOG, is the leading tech company that decided to invest a lot of resources and funding in artificial intelligence. So much so, that the WSJ recently announced that AI is central to Google's future. Not surprisingly, Google has been dealing with different challenges concerning its top AI executives and researchers. Activists shareholders are also showing interest in this. Recently, there is a rise in shareholder proposals calling on boards to ensure proper AI governance.

Facial recognition technology has been banned by multiple US cities, including Portland, Boston, and San Francisco. Besides the very real risk of the tech being biased against minorities, the technology also carries with it an uneasy sense that we're creeping towards a surveillance state. Despite these concerns, though, work to improve facial recognition tech is still forging ahead, with both private companies and governments looking to harness its potential for military, law enforcement, or profit-seeking applications. One such company is an Israeli startup called AnyVision Interactive Technologies. AnyVision is looking to kick facial recognition up a notch by employing drones for image capture.

Here are some of the most significant themes we see as we look toward 2021. Some of these are emerging topics and others are developments on existing concepts, but all of them will inform our thinking in the coming year. MLOps attempts to bridge the gap between Machine Learning (ML) applications and the CI/CD pipelines that have become standard practice. ML presents a problem for CI/CD for several reasons. The data that powers ML applications is as important as code, making version control difficult; outputs are probabilistic rather than deterministic, making testing difficult; training a model is processor intensive and time consuming, making rapid build/deploy cycles difficult.

With the advent of digital technology, it is more common that committed crimes or legal disputes involve some form of speech recording where the identity of a speaker is questioned [1]. In face of this situation, the field of forensic speaker identification has been looking to shed light on the problem by quantifying how much a speech recording belongs to a particular person in relation to a population. In this work, we explore the use of speech embeddings obtained by training a CNN using the triplet loss. In particular, we focus on the Spanish language which has not been extensively studies. We propose extracting the embeddings from speech spectrograms samples, then explore several configurations of such spectrograms, and finally, quantify the embeddings quality. We also show some limitations of our data setting which is predominantly composed by male speakers. At the end, we propose two approaches to calculate the Likelihood Radio given out speech embeddings and we show that triplet loss is a good alternative to create speech embeddings for forensic speaker identification.

Darryl Richardson was delighted when he landed a job as a "picker" at the Amazon warehouse in Bessemer, Alabama. "I thought, 'Wow, I'm going to work for Amazon, work for the richest man around," he said. "I thought it would be a nice facility that would treat you right." Richardson, a sturdily built 51-year-old with a short, charcoal beard, took a job at the gargantuan warehouse after the auto parts plant where he worked for nine years closed. Now he is strongly supporting the ambitious effort to unionize its 5,800 workers because, he says, the job is so demanding and working for Amazon has fallen far below his expectations. Last August, five months after the warehouse opened, Richardson began pushing for a union in what is not only the first effort to organize an entire Amazon warehouse in the United States, but also the biggest private-sector union drive in the south in years. "I thought the opportunities for moving up would be better. I thought safety at the plant would be better," Richardson said. "And when it comes to letting people go for no reason – job security – I thought it would be different."

Synthetic media technologies are rapidly advancing, making it easier to generate nonveridical media that look and sound increasingly realistic. So-called "deepfakes" (owing to their reliance on deep learning) often present a person saying or doing something they have not said or done. The proliferation of deepfakesa creates a new challenge to the trustworthiness of visual experience, and has already created negative consequences such as nonconsensual pornography,11 political disinformation,19 and financial fraud.3 Deepfakes can harm viewers by deceiving or intimidating, harm subjects by causing reputational damage, and harm society by undermining societal values such as trust in institutions.7 What can be done to mitigate these harms?

A United Kingdom court dropped Epic Games' suit against Apple and its request to make the tech giant reinstate its popular video game "Fortnite" into the App Store. In the meantime, sorry, Apple users, you are still shut out from playing "Fortnite" with your friends who are blasting away on PlayStations and Xboxes, for instance. The U.S., where Epic Games and Apple are headquartered, would be "the appropriate forum" for the cases to be tried, said Judge Justice Roth of the Competition Appeal Tribunal in a ruling Monday. This legal battle, which began in August 2020 when Epic offered a direct payment method for Fortnite mobile players, spans the globe. Last week, Epic Games filed an antitrust complaint against Apple in the European Union.

With the widespread deployment of large-scale prediction systems in high-stakes domains, e.g., face recognition, criminal justice, etc., disparity on prediction accuracy between different demographic subgroups has called for fundamental understanding on the source of such disparity and algorithmic intervention to mitigate it. In this paper, we study the accuracy disparity problem in regression. To begin with, we first propose an error decomposition theorem, which decomposes the accuracy disparity into the distance between marginal label distributions and the distance between conditional representations, to help explain why such accuracy disparity appears in practice. Motivated by this error decomposition and the general idea of distribution alignment with statistical distances, we then propose an algorithm to reduce this disparity, and analyze its game-theoretic optima of the proposed objective functions. To corroborate our theoretical findings, we also conduct experiments on five benchmark datasets. The experimental results suggest that our proposed algorithms can effectively mitigate accuracy disparity while maintaining the predictive power of the regression models.

Online convex optimization is a framework where a learner sequentially queries an external data source in order to arrive at the optimal solution of a convex function. The paradigm has gained significant popularity recently thanks to its scalability in large-scale optimization and machine learning. The repeated interactions, however, expose the learner to privacy risks from eavesdropping adversary that observe the submitted queries. In this paper, we study how to optimally obfuscate the learner's queries in first-order online convex optimization, so that their learned optimal value is provably difficult to estimate for the eavesdropping adversary. We consider two formulations of learner privacy: a Bayesian formulation in which the convex function is drawn randomly, and a minimax formulation in which the function is fixed and the adversary's probability of error is measured with respect to a minimax criterion. We show that, if the learner wants to ensure the probability of accurate prediction by the adversary be kept below $1/L$, then the overhead in query complexity is additive in $L$ in the minimax formulation, but multiplicative in $L$ in the Bayesian formulation. Compared to existing learner-private sequential learning models with binary feedback, our results apply to the significantly richer family of general convex functions with full-gradient feedback. Our proofs are largely enabled by tools from the theory of Dirichlet processes, as well as more sophisticated lines of analysis aimed at measuring the amount of information leakage under a full-gradient oracle.

Corruption continues to be one of the biggest societal challenges of our time. New hope is placed in Artificial Intelligence (AI) to serve as an unbiased anti-corruption agent. Ever more available (open) government data paired with unprecedented performance of such algorithms render AI the next frontier in anti-corruption. Summarizing existing efforts to use AI-based anti-corruption tools (AI-ACT), we introduce a conceptual framework to advance research and policy. It outlines why AI presents a unique tool for top-down and bottom-up anti-corruption approaches. For both approaches, we outline in detail how AI-ACT present different potentials and pitfalls for (a) input data, (b) algorithmic design, and (c) institutional implementation. Finally, we venture a look into the future and flesh out key questions that need to be addressed to develop AI-ACT while considering citizens' views, hence putting "society in the loop".