Summarization systems make numerous "decisions" about summary properties during inference, e.g. degree of copying, specificity and length of outputs, etc. However, these are implicitly encoded within model parameters and specific styles cannot be enforced. To address this, we introduce HydraSum, a new summarization architecture that extends the single decoder framework of current models to a mixture-of-experts version with multiple decoders. We show that HydraSum's multiple decoders automatically learn contrasting summary styles when trained under the standard training objective without any extra supervision. Through experiments on three summarization datasets (CNN, Newsroom and XSum), we show that HydraSum provides a simple mechanism to obtain stylistically-diverse summaries by sampling from either individual decoders or their mixtures, outperforming baseline models. Finally, we demonstrate that a small modification to the gating strategy during training can enforce an even stricter style partitioning, e.g. high- vs low-abstractiveness or high- vs low-specificity, allowing users to sample from a larger area in the generation space and vary summary styles along multiple dimensions.

Between 2019 and 2022, Greek media attention has been attracted by a rather unusually high number of femicide cases which have been trending for several weeks up to months in the public debate and one of the contributing factors is the feedback loop between traditional media and social media. In this paper we are investigating the use of the term "\{gamma}\u{psion}{\nu}{\alpha}{\iota}\k{appa}{\omicron}\k{appa}{\tau}{\omicron}{\nu}{\iota}{\alpha}" (femicide) in Greek speaking twitter. More specifically, we approach the problem from a stance detection perspective, aiming to automatically identify user position with regards to the feministic semantics of the word. We also discuss findings from an identity analysis perspective and intercorrelations with hate speech that have been identified in the collected corpus of tweets.

The office of Texas State Attorney General Ken Paxton announced on Thursday that it has filed a lawsuit against Google over the company's alleged years-long practices to capture and use of biometric data from, "millions of Texans without properly obtaining their informed consent to do so." This is allegedly a violation of the state's Capture or Use of Biometric Identifier Act of 2009. The AG argues that Google used features in its Photos and Assistant apps, as well as through Nest Hub Max hardware, to scan and store the facial and voice data without first acquiring user consent. Furthermore, Paxton alleges, Google then leveraged that data for commercial gain by using it to train the company's machine learning algorithms. "Google's indiscriminate collection of the personal information of Texans, including very sensitive information like biometric identifiers, will not be tolerated," Texas Attorney General Ken Paxton said in the Thursday press release.

The Texas attorney general sued Alphabet Google on Thursday, alleging the search giant violated state laws by collecting biometric data on face and voice features without seeking the full consent of users. Texas alleged Google's data-collection practices stretched back to 2015 and affected millions of the state's residents, according to a complaint filed in state district court in Midland County, Texas. A weekly digest of tech reviews, headlines, columns and your questions answered by WSJ's Personal Tech gurus. "Google's indiscriminate collection of the personal information of Texans, including very sensitive information like biometric identifiers, will not be tolerated," Texas Attorney General Ken Paxton said. "I will continue to fight Big Tech to ensure the privacy and security of all Texans."

The EU's proposed Artificial Intelligence Act plans to restrict open-source AI. The proposed – and still debated – Artificial Intelligence Act (AIA) from the EU touches upon the regulation of open-source AI. But enforcing strict restrictions on the sharing and distribution of open-source general-purpose AI (GPAI) is a completely retrograde step. It is like rewinding the world back by 30 years. Open-source culture is the only reason why mankind was able to progress technology at such a light speed. Only recently AI researchers were able to embrace sharing their code for more transparency and verification but putting constraints on this movement will damage the cultural progress the scientific community has made.

A recently passed law in New York City requires audits for bias in AI-based hiring systems. AI systems fail frequently, and bias is often to blame. A recent sampling of headlines features sociological bias in generated images, a chatbot, and a virtual rapper. These examples of denigration and stereotyping are troubling and harmful, but what happens when the same types of systems are used in more sensitive applications? Leading scientific publications assert that algorithms used in healthcare in the U.S. diverted care away from millions of black people.

Orgaimi Inc., which makes a client intelligence platform powered by artificial intelligence, said Wednesday that it had hired a former Hogan Lovells head of innovation and digital to be a market evangelist.

A $1000 CAD (or equivalent in your country's currency) work from home allowance to make your home setup perfect for you A lifestyle spending account for employees to receive reimbursement for eligible expenses related to wellness, lifestyle and productivity $2500 CAD (or equivalent in your country's currency) per year At BenchSci, we're committed to cultivating an inspiring, inclusive, and equitable work environment for high performing, ego-free, self-starting individuals with a growth mindset, who enjoy the challenge of solving hard problems. We recognize that everyone here is a person first and an employee second. We want people to feel cared for and supported to bring the best versions of themselves to work and help the company achieve its mission. We believe culture is critical to success and invest accordingly. We live and promote our FASTT values of Focused, Advancement with Speed, Tenacity, and Transparency. We work hard to maintain an engaging, supportive environment where everyone can do their best work. To learn more, read our culture deck.

With the update, Pegasystems also targeted business users with new AI capabilities for its Pega Platform automation system, including Developer Assistant panel, an application that provides recommendations to developers; Prediction Widget, which shows users how AI-powered predictions are made; and Pega Robot Studio, which enables users to build and deploy bots to automate repetitive tasks. The vendor also focused on its core customer engagement technology with new capabilities such as Pega Customer Data Connectors that link Pega Hub with customer data platforms such as Adobe and Celebrus, and Impact Analyzer for monitoring business goals and AI models in rea time. While competitors such as Salesforce with its Einstein AI system, Verint and Appian -- which was awarded a $2b judgement against Pegasystems in a theft of trade secrets lawsuit that Pega is appealing -- offer similar automation products, Pegasystems still has some room to compete, said Predrag Jakovljevic, an analyst at Technology Evaluation Centers. "They can compete by making it easier to use, and to find their strongholds in certain industries like insurance and financial institutions," Jakovljevic said, adding that Pegasystems has long been strong in decisioning, next best action and case management in applications such as insurance claims and mortgage forms.

Deidentification seeks to anonymize textual data prior to distribution. Automatic deidentification primarily uses supervised named entity recognition from human-labeled data points. We propose an unsupervised deidentification method that masks words that leak personally-identifying information. The approach utilizes a specially trained reidentification model to identify individuals from redacted personal documents. Motivated by K-anonymity based privacy, we generate redactions that ensure a minimum reidentification rank for the correct profile of the document. To evaluate this approach, we consider the task of deidentifying Wikipedia Biographies, and evaluate using an adversarial reidentification metric. Compared to a set of unsupervised baselines, our approach deidentifies documents more completely while removing fewer words. Qualitatively, we see that the approach eliminates many identifying aspects that would fall outside of the common named entity based approach.