Waymo has sued the California Department of Motor Vehicles. In a case first reported by The Los Angeles Times, the Alphabet subsidiary filed a complaint with the Sacramento County Superior Court on January 21st to prevent the agency from disclosing what it believes to be trade secrets. At the center of the lawsuit is a public records request an unidentified party made to obtain Waymo's driverless deployment application. Before sharing the requested documents, the DMV allowed the company to redact any sections it believed would reveal its trade secrets, including questions that were asked by the agency. When the DMV eventually forwarded the package to the requester, that individual or group challenged the redactions.

First, financial regulators need to ensure that regulatory oversight delivers on the inclusion and intermediation-enhancing benefits of digital finance without compromising traditional regulatory goals such as financial stability, adequate competition, consumer protection and market integrity. Second, there is a pressing need for a system of data governance that allows consumers and business to exercise control over their data through the granting and withholding of consent to the use and transfer of their data. Developing a user-friendly granular consent-based data governance system with low transaction costs is a challenge that, when successfully addressed, will promote the development of virtual banking worldwide. Hong Kong SAR offers one example of an integrated regulatory framework for virtual banks. The licensing and regulatory regime – also applicable to incumbent banks – aims to manage the full spectrum of risks arising from any source, including the ownership structure, without compromising development objectives that often rest on technological innovation.

This paper demonstrates the systematic use of combinatorial coverage for selecting and characterizing test and training sets for machine learning models. The presented work adapts combinatorial interaction testing, which has been successfully leveraged in identifying faults in software testing, to characterize data used in machine learning. The MNIST hand-written digits data is used to demonstrate that combinatorial coverage can be used to select test sets that stress machine learning model performance, to select training sets that lead to robust model performance, and to select data for fine-tuning models to new domains. Thus, the results posit combinatorial coverage as a holistic approach to training and testing for machine learning. In contrast to prior work which has focused on the use of coverage in regard to the internal of neural networks, this paper considers coverage over simple features derived from inputs and outputs. Thus, this paper addresses the case where the supplier of test and training sets for machine learning models does not have intellectual property rights to the models themselves. Finally, the paper addresses prior criticism of combinatorial coverage and provides a rebuttal which advocates the use of coverage metrics in machine learning applications.

Did you miss a session from the Future of Work Summit? Large language models (LLMs) such as OpenAI's GPT-3, which can "write" sentences that read nearly like they were written by a human, can be prompted to perform a range of writing tasks given only a few examples of the tasks. For example, LLMs have been used to create marketing materials and video game levels in addition to recipes, poetry, and movie scripts. But because LLMs learn to write from examples taken from sometimes toxic communities, they can fall victim to parroting misinformation, sexism, ageism, racism, and conspiracies. Efforts have been made to combat toxicity in LLMs -- with mixed results.

Tech companies in the U.S. and the U.K. haven't done enough to prevent bias in artificial intelligence algorithms, according to a new survey from Data Robot. These same organizations are already feeling the impact of this problem as well in the form of lost customers and lost revenue. DataRobot surveyed more than 350 U.S. and U.K.-based technology leaders to understand how organizations are identifying and mitigating instances of AI bias. Survey respondents included CIOs, IT directors, IT managers, data scientists and development leads who use or plan to use AI. The research was conducted in collaboration with the World Economic Forum and global academic leaders.

Protected attributes are often presented as categorical features that need to be encoded before feeding them into a machine learning algorithm. Encoding these attributes is paramount as they determine the way the algorithm will learn from the data. Categorical feature encoding has a direct impact on the model performance and fairness. In this work, we compare the accuracy and fairness implications of the two most well-known encoders: one-hot encoding and target encoding. We distinguish between two types of induced bias that can arise while using these encodings and can lead to unfair models. The first type, irreducible bias, is due to direct group category discrimination and a second type, reducible bias, is due to large variance in less statistically represented groups. We take a deeper look into how regularization methods for target encoding can improve the induced bias while encoding categorical features. Furthermore, we tackle the problem of intersectional fairness that arises when mixing two protected categorical features leading to higher cardinality. This practice is a powerful feature engineering technique used for boosting model performance. We study its implications on fairness as it can increase both types of induced bias

Federated learning (FL) is vulnerable to heterogeneously distributed data, since a common global model in FL may not adapt to the heterogeneous data distribution of each user. To counter this issue, personalized FL (PFL) was proposed to produce dedicated local models for each individual user. However, PFL is far from its maturity, because existing PFL solutions either demonstrate unsatisfactory generalization towards different model architectures or cost enormous extra computation and memory. In this work, we propose federated learning with personalized sparse mask (FedSpa), a novel PFL scheme that employs personalized sparse masks to customize sparse local models on the edge. Instead of training an intact (or dense) PFL model, FedSpa only maintains a fixed number of active parameters throughout training (aka sparse-to-sparse training), which enables users' models to achieve personalization with cheap communication, computation, and memory cost. We theoretically show that the iterates obtained by FedSpa converge to the local minimizer of the formulated SPFL problem at rate of $\mathcal{O}(\frac{1}{\sqrt{T}})$. Comprehensive experiments demonstrate that FedSpa significantly saves communication and computation costs, while simultaneously achieves higher model accuracy and faster convergence speed against several state-of-the-art PFL methods.

A new series of patents by Facebook's parent company Meta suggest possible plans from the company to capture users' behavioral biometrics data. Specifically, the patents mention pupil movements, body poses, and crumpled noses, which the company would use to make digital avatars realistically animated. The patents were reviewed by The Financial Times, according to which Meta also intends to use the biometric data to provide hyper-targeted advertising and sponsored content. In fact, one of the patents analyzed by the publication was granted to Meta by the United States Patent and Trademark Office (USPTO) earlier this month and refers to the tracking of users' facial expressions through a virtual reality headset to "adapt media content" based on those responses. A separate patent describes an avatar personalization engine capable of creating a 3D avatar of a user based on biometrics collected from a submitted photo.

Pimloc, a UK computer vision startup that's sharpened its business pitch to sell an AI service for quickly anonymizing video -- automating the blurring of faces or licence plates, along with a suite of other visual search services -- has grabbed another chunk of seed funding: Announcing a raise of $7.5M, led by Zetta Venture Partners, with participation from existing investors Amadeus Capital Partners and Speedinvest. The startup raised a $1.8M seed, back in October 2020, but says the new funds will be used to scale the business across Europe and the U.S., tracking the spread of data legislation and the evolution of public opinion around the privacy risks of biometrics -- pointing, for example, to the privacy backlash around Clearview AI. As well as building out its sales, marketing and R&D teams, Pimloc says the funding will be used to expand its product roadmap with a focus on video privacy and compliance. The business need it's targeting focuses on growing use of visual AI in industries like retail, warehousing and industrial factory settings -- for use cases like safety and efficiency. However the rise of AI-powered workplace surveillance tools create privacy risks for workers which could create legal and reputational risks for companies that deploy remote biometrics.

Whether you've used social media, a navigation app or a picture filter, chances are that Artificial Intelligence (AI) has impacted you. It's not just you — AI is impacting human rights worldwide, and this course will inform and educate you on how your rights are affected by AI, and how you can be empowered to guard these rights. UNESCO and UNITAR jointly launched a new, short online learning course on AI and Human Rights for youths aged 16 to 24. Experts break down complex concepts about AI into straight forward activities built around our daily technology interactions. The course focuses on how freedom of expression, right to privacy and the right to equality are impacted using AI.