How it works: Protective tools like Glaze and Nightshade change enough pixels to affect an image, so if it's scraped up by AI models, they see it as something it's not. LightShed essentially works by spotting just the "poison" on poisoned images. To be clear, the researchers behind it aren't trying to steal artists' work. They just don't want people to get a false sense of security. The "Big, Beautiful Bill" that President Donald Trump signed into law on July 4 was chock full of controversial policies.

Ukrainian President Volodymyr Zelenskyy has opened a conference in Italy on rebuilding his war-battered country, as it comes under relentless ground and air attacks from Russia. The Rome gathering will see the Ukrainian leader hold a flurry of meetings on Thursday, including a video call with leaders from about 30 countries in the so-called "coalition of the willing", as he seeks to secure financing to bolster his country's air defence systems, which were this week strained by Russia's largest missile and drone attack in more than three years of war. The United Kingdom and France are spearheading talks among the coalition on how to support a possible ceasefire in Ukraine, including potentially deploying peacekeeping forces to police any future peace agreement with Russia. This week, the office of UK Prime Minister Keir Starmer said the call would cover "stepping up support for Ukraine and further increasing pressure on Russia". The success of the coalition's operation hinges on United States backup with airpower or other military assistance, but the administration of President Donald Trump has made no public commitment to provide support. Amid growing uncertainty about US commitment to Kyiv's defence, despite Trump's recent U-turn on pausing critical weapons deliveries, Zelenskyy had a "substantive" meeting with Trump's Ukraine envoy, Keith Kellogg, on Wednesday.

Elon Musk's artificial intelligence company xAI has come under fire after its chatbot Grok stirred controversy with anti-Semitic responses to questions posed by users – just weeks after Musk said he would rebuild it because he felt it was too politically correct. On Friday last week, Musk announced that xAI had made significant improvements to Grok, promising a major upgrade "within a few days". Online tech news site The Verge reported that, by Sunday evening, xAI had already added new lines to Grok's publicly posted system prompts. By Tuesday, Grok had drawn widespread backlash after generating inflammatory responses – including anti-Semitic comments. One Grok user asking the question, "which 20th-century figure would be best suited to deal with this problem (anti-white hate)", received the anti-Semitic response: "To deal with anti-white hate? Here's what we know about the Grok chatbot and the controversies it has caused. Grok, a chatbot created by xAI – the AI company Elon Musk ...

The clash between billionaire Elon Musk's xAI empire and European officials is intensifying with leaders in Poland and Germany calling for more aggressive action against the company. German lawmaker Ralf Stegner, responding to antisemitic comments that xAI's chatbot Grok made Tuesday on Musk's social media platform, X, said the posts "must not be tolerated under any circumstances" and called for sanctions in an interview with the German newspaper Handelsblatt. Poland's government separately urged the European Union to investigate and possibly fine xAI following lewd comments made by Grok about the country's politicians. The European Union is already investigating Musk's social media platform under a relatively new content-moderation policy known as the Digital Services Act and had been weighing a fine ahead of its summer recess in August. The regulator is reportedly considering calculating the fine by including revenue from Musk's other businesses, including SpaceX and Neuralink, an approach that would significantly increase the potential penalties.

Ukraine's capital Kyiv is again under a massive overnight Russian drone attack, local officials say, with at least 10 people reported injured and fires burning across the city. Authorities in Kyiv say drone wreckage has hit the roof of a residential building in the central Shevchenkivskyi district. Footage on social media, as yet unverified by the BBC, shows explosions in the night sky, as air defence units begin repelling the attack. Ukraine's military has also warned of a threat of a ballistic missile attack. Last night, Ukraine reported the biggest ever aerial attack from Russia, after 728 drones and 13 cruise or ballistic missiles struck cities around the country in multiple waves.

Diffusion models have emerged as a powerful class of generative models, achieving state-of-the-art results in continuous data domains such as image and video generation. Their core mechanism involves a forward diffusion process that gradually transforms structured data into a Gaussian-like distribution, followed by a learned reverse process to reconstruct the data. While successful in continuous modalities, applying this framework to discrete data-particularly natural language-remains challenging due to token dependency complexities and the lack of a defined generation order.This thesis investigates the feasibility and performance of discrete diffusion models for natural language generation. Specifically, we evaluate the Discrete Denoising Diffusion Probabilistic Model (D3PM) and compare it with traditional autoregressive (AR) language models. To assess generative performance, we use Bits Per Token (BPT), Negative Log-Likelihood (NLL), Perplexity (PPL), and Batch Processing Speed. Results show the best-performing D3PM model achieves a BPT of 5.72, with a mean of 8.05. The AR model outperforms in compression with a lower mean BPT of 4.59, but D3PM achieves higher processing speed, reaching up to 3.97 batches per sec., indicating potential for parallel generation.All evaluations were conducted under consistent conditions-generating 100,000 tokens per model with a fixed batch size of four-for fair comparison. This research presents a detailed analysis of diffusion-based vs. autoregressive models, highlighting trade-offs in generative quality and efficiency. Findings emphasize both the promise and limitations of diffusion models for discrete data, supporting future work in non-autoregressive language generation.

Information technology is in the midst of a revolution in which omnipresent data collection and machine learning are impacting the human world as never before. The word "intelligence" is being used as a North Star for the development of this technology, with human cognition viewed as a baseline. This view neglects the fact that humans are social animals, and that much of our intelligence is social and cultural in origin. A related issue is that the current view treats the social consequences of technology as an afterthought. The path forward is not merely more data and compute, and not merely more attention paid to cognitive or symbolic representations, but a thorough blending of economic and social concepts with computational and inferential concepts, in the service of system-level designs in which social welfare is a first-class citizen, and with the aspiration that a new human-centric engineering field will emerge.

Providing generalization guarantees for stochastic optimization algorithms is a major challenge in modern learning theory. Recently, several studies highlighted the impact of the geometry of training trajectories on the generalization error, both theoretically and empirically. Among these works, a series of topological generalization bounds have been proposed, relating the generalization error to notions of topological complexity that stem from topological data analysis (TDA). Despite their empirical success, these bounds rely on intricate information-theoretic (IT) terms that can be bounded in specific cases but remain intractable for practical algorithms (such as ADAM), potentially reducing the relevance of the derived bounds. In this paper, we seek to formulate comprehensive and interpretable topological generalization bounds free of intractable mutual information terms. To this end, we introduce a novel learning theoretic framework that departs from the existing strategies via proof techniques rooted in algorithmic stability. By extending an existing notion of \textit{hypothesis set stability}, to \textit{trajectory stability}, we prove that the generalization error of trajectory-stable algorithms can be upper bounded in terms of (i) TDA quantities describing the complexity of the trajectory of the optimizer in the parameter space, and (ii) the trajectory stability parameter of the algorithm. Through a series of experimental evaluations, we demonstrate that the TDA terms in the bound are of great importance, especially as the number of training samples grows. This ultimately forms an explanation of the empirical success of the topological generalization bounds.

Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks -- re-identification, attribute inference, and data reconstruction -- are both overly pessimistic and inconsistent. In this work, we use the hypothesis-testing interpretation of DP ($f$-DP), and determine that bounds on attack success can take the same unified form across re-identification, attribute inference, and data reconstruction risks. Our unified bounds are (1) consistent across a multitude of attack settings, and (2) tunable, enabling practitioners to evaluate risk with respect to arbitrary (including worst-case) levels of baseline risk. Empirically, our results are tighter than prior methods using $\varepsilon$-DP, Rényi DP, and concentrated DP. As a result, calibrating noise using our bounds can reduce the required noise by 20% at the same risk level, which yields, e.g., more than 15pp accuracy increase in a text classification task. Overall, this unifying perspective provides a principled framework for interpreting and calibrating the degree of protection in DP against specific levels of re-identification, attribute inference, or data reconstruction risk.

This paper presents a general framework for conducting efficient inference on parameters derived from unstructured data, which include text, images, audio, and video. Economists have long used unstructured data by first extracting low-dimensional structured features (e.g., the topic or sentiment of a text), since the raw data are too high-dimensional and uninterpretable to include directly in empirical analyses. The rise of deep neural networks has accelerated this practice by greatly reducing the costs of extracting structured data at scale, but neural networks do not make generically unbiased predictions. This potentially propagates bias to the downstream estimators that incorporate imputed structured data, and the availability of different off-the-shelf neural networks with different biases moreover raises p-hacking concerns. To address these challenges, we reframe inference with unstructured data as a problem of missing structured data, where structured variables are imputed from high-dimensional unstructured inputs. This perspective allows us to apply classic results from semiparametric inference, leading to estimators that are valid, efficient, and robust. We formalize this approach with MAR-S, a framework that unifies and extends existing methods for debiased inference using machine learning predictions, connecting them to familiar problems such as causal inference. Within this framework, we develop robust and efficient estimators for both descriptive and causal estimands and address challenges like inference with aggregated and transformed missing structured data-a common scenario that is not covered by existing work. These methods-and the accompanying implementation package-provide economists with accessible tools for constructing unbiased estimators using unstructured data in a wide range of applications, as we demonstrate by re-analyzing several influential studies.