The Hidden Costs of AI 'Optimization': When Efficiency Erodes Ethics
A new wave of machine learning research, published today on arXiv CS. LG, highlights the persistent technical hurdles in building reliable and ethical AI systems....
Read article →
Leila Grant
Opinion Columnist
Technology & Society
Leila Grant writes about agency, access and the social choices embedded in technical systems. Her columns consider whose interests a technology serves and whose experience is absent from the debate. She combines an interest in possibility with close attention to power.
Articles by Leila
433 published articles · Showing 61–80
Technology's Dual Grasp: Reshaping Industrial Metals and the Fabric of Life Itself
The fundamental components of our world, from the steel in our infrastructure to the biological processes of life, are being rapidly reconfigured by technological advancements. This week, two distinct developments underscore this acceleration: Boston Metal secured $75 million to ...
Read article →The Invisible Hand of AI: When Machines Decide What Our Words Should Be
A new research paper introduces FormalASR, an end-to-end system designed to convert spoken Chinese directly into "formal text," bypassing the nuanced irregularities of natural speech arXiv CS. AI....
Read article →New Research Challenges Algorithmic Bias in Privacy-Preserving AI
Federated learning, often presented as a solution for privacy in AI development, carries a subtle but significant risk: it can amplify existing societal inequalities. New research published on arXiv CS....
Read article →AI Takes Autonomy: Algorithms Now Tune Systems and Judge Code, Raising Questions of Control and Transparency
Imagine a PostgreSQL database, a complex system designed to serve myriad users, its very heartbeat – its parameters – once meticulously tweaked by seasoned engineers guided by expert documentation. Now, an emerging class of AI systems is poised to take over this delicate work, sh...
Read article →Examining the Cracks: New Research Reveals Limitations in AI Safety Mechanisms
Large Language Models, once heralded as controlled and alignable, are proving deeply vulnerable to manipulation. New research published today on arXiv CS....
Read article →The Promise of Leaner AI: Who Truly Profits from Compression?
A flurry of new research, published today on arXiv CS. AI, signals significant progress in AI model quantization and compression....
Read article →New AI Research Promises Medical Breakthroughs, Raises Urgent Questions of Access and Ethics
On May 20, 2026, a flurry of new machine learning research published on arXiv CS. LG unveiled potential AI applications across critical areas of healthcare, from predicting heart failure readmission to diagnosing rare immune disorders and assessing malaria severity....
Read article →The Invisible Hands Guiding Our Robots: New 'Crowdsourcing' Model Raises Ethical Red Flags
Somewhere, a worker taps a smartphone screen, guiding a virtual robot arm to grasp an object. This mundane action, repeated thousands of times, is the bedrock of 'COBALT,' a new platform designed to crowdsource the vast datasets needed for advanced robotic manipulation arXiv CS....
Read article →The Unstable Foundations of AI: Whose Truth Do We Build Upon?
The quiet certainty that underpins much of AI development is being shaken. New research reveals not only a fundamental "Western-centric default" in AI safety frameworks, creating "significant vulnerabilities" in the Global South arXiv CS....
Read article →AI Research Advances Reasoning and Autonomous Systems: The Unanswered Questions
On May 19, 2026, two significant research papers published on arXiv detailed advancements in artificial intelligence, pushing the capabilities of AI reasoning in agriculture and improving the foundational speed of autonomous visual-inertial systems arXiv CS. AI arXiv CS....
Read article →New arXiv Papers Advance Machine Learning Foundations: A Call for Proactive Ethics
A significant cluster of seven new research papers appeared on arXiv CS. LG today, signaling fresh advancements in the theoretical underpinnings of machine learning and statistical methods....
Read article →Whose 'Transformation'? AI's New Velocity and the Cost of Obedience
New research published today on arXiv details significant advancements in AI's capacity to accelerate scientific and engineering simulations, promising orders-of-magnitude efficiency gains across industries arXiv CS. AI....
Read article →Generative AI Aims to 'Visualize the Invisible' in Brain Activity: A New Frontier for Mental Privacy?
A new research paper introduces Generative Visual Grounding (GVG), a framework designed to interpret visually-evoked brain activity (EEG) with unprecedented detail using Multi-modal Large Language Models (MLLMs). Published in arXiv CS....
Read article →Musk's OpenAI Lawsuit Loss Reveals Deeper Battle Over AI's Public Purpose
The legal battle over OpenAI's founding principles concluded today with Elon Musk's loss in his suit against CEO Sam Altman and President Greg Brockman. Musk alleged they deceived him regarding the company's non-profit status MIT Tech Review....
Read article →OpenAI Pushes 'Provenance' Tools as Experts Demand Deeper Questions on AI Ethics
A new initiative from OpenAI aims to establish clearer content provenance through Content Credentials, SynthID, and a dedicated verification tool, announced today, May 19, 2026 OpenAI Blog. The company asserts these tools will foster a "safer, more transparent AI ecosystem....
Read article →The Price of Connectivity: How Tech's Power Hunger Lights Up Your Utility Bill
The latest blockbuster deal in the energy sector promises a future of higher utility bills for ordinary people. NextEra's proposed merger with Dominion is not about grid efficiency or green energy innovation....
Read article →AI's Erasure Problem: New Research Tackles Harmful LLM Outputs, But Who Decides What Stays and What Goes?
When a machine is designed to forget, the critical question becomes: who decides what memories are erased, and whose history is silenced? New research published on arXiv explores the complex challenge of "unlearning" harmful or sensitive information from large language models (LL...
Read article →The Silent Cost of AI Efficiency: Who Profits When Models Get 'Optimized'?
Today, a flurry of new research, published across a dozen papers on arXiv, signals an accelerating, focused effort to optimize large language models (LLMs) and other AI systems for peak efficiency. This intense drive, manifest in new methods for hardware-aware architecture search...
Read article →New Research Exposes AI's Fragile Foundations, Demanding Accountability
The ongoing struggle to build truly trustworthy artificial intelligence systems continues, with new research highlighting critical gaps in how we evaluate these complex models. Recent preprints on arXiv CS....
Read article →