We examine the ability of large language models (LLMs) to generate salient (interesting) negative statements about real-world entities; an emerging research topic of the last few years. We probe the LLMs using zero- and k-shot unconstrained probes, and compare with traditional methods for negation generation, i.e., pattern-based textual extractions and knowledge-graph-based inferences, as well as crowdsourced gold statements. We measure the correctness and salience of the generated lists about subjects from different domains. Our evaluation shows that guided probes do in fact improve the quality of generated negatives, compared to the zero-shot variant. Nevertheless, using both prompts, LLMs still struggle with the notion of factuality of negatives, frequently generating many ambiguous statements, or statements with negative keywords but a positive meaning.

Object-centric learning aims to represent visual data with a set of object entities (a.k.a. slots), providing structured representations that enable systematic generalization. Leveraging advanced architectures like Transformers, recent approaches have made significant progress in unsupervised object discovery. In addition, slot-based representations hold great potential for generative modeling, such as controllable image generation and object manipulation in image editing. However, current slot-based methods often produce blurry images and distorted objects, exhibiting poor generative modeling capabilities. In this paper, we focus on improving slot-to-image decoding, a crucial aspect for high-quality visual generation. We introduce SlotDiffusion -- an object-centric Latent Diffusion Model (LDM) designed for both image and video data. Thanks to the powerful modeling capacity of LDMs, SlotDiffusion surpasses previous slot models in unsupervised object segmentation and visual generation across six datasets. Furthermore, our learned object features can be utilized by existing object-centric dynamics models, improving video prediction quality and downstream temporal reasoning tasks. Finally, we demonstrate the scalability of SlotDiffusion to unconstrained real-world datasets such as PASCAL VOC and COCO, when integrated with self-supervised pre-trained image encoders.

In recent years, the burgeoning interest in diffusion models has led to significant advances in image and speech generation. Nevertheless, the direct synthesis of music waveforms from unrestricted textual prompts remains a relatively underexplored domain. In response to this lacuna, this paper introduces a pioneering contribution in the form of a text-to-waveform music generation model, underpinned by the utilization of diffusion models. Our methodology hinges on the innovative incorporation of free-form textual prompts as conditional factors to guide the waveform generation process within the diffusion model framework. Addressing the challenge of limited text-music parallel data, we undertake the creation of a dataset by harnessing web resources, a task facilitated by weak supervision techniques. Furthermore, a rigorous empirical inquiry is undertaken to contrast the efficacy of two distinct prompt formats for text conditioning, namely, music tags and unconstrained textual descriptions. The outcomes of this comparative analysis affirm the superior performance of our proposed model in terms of enhancing text-music relevance. Finally, our work culminates in a demonstrative exhibition of the excellent capabilities of our model in text-to-music generation. We further demonstrate that our generated music in the waveform domain outperforms previous works by a large margin in terms of diversity, quality, and text-music relevance.

Speaking in regard to everything from AI-powered cyber attacks, to the risk of malfunctioning AI, how AI can spread misinformation, and even the interaction between AI and nuclear weapons, Mr Guterres said: "Without action to address these risks, we are derelict in our responsibilities to present and future generations."

The lawsuit is the latest salvo in the ongoing debate over how AI tools should be trained and whether the companies behind them owe anything to the original creators of the training data. Large language models are generally trained on billions of sentences of text pulled from the internet, including news stories, Wikipedia and comments on social media sites. OpenAI and other AI companies such as Google and Microsoft do not say specifically what data they use, but AI critics have long suspected that it includes well-known collections of pirated books that have circulated online for years.

Where does art begin and end? The question is at the center of a debate that roiled X (formerly known as Twitter) this month after an AI-generated image of a medieval village, titled Spiral Town, went viral. "I stole this from someone on Twitter who stole this from someone on Reddit," a user named @deepfates posted. "shout out to all of humanity [...] who contributed training data." Generative AI moves at light speed, a pace so unpredictable that sometimes even I struggle to keep up.

Kurt "The Cyberguy" Knutsson explains how the new Apollo humanoid can potentially be the end of jobs as we know it. Are we living in the future? Are the robots taking over? No … but for Austin-based startup Apptronik, robots are being developed and are here to stay. Meet Apollo: Apptronik's latest "general purpose humanoid robot" powered by artificial intelligence (AI).

The'America's Got Talent' judge told Fox News Digital why he doesn't like AI technology in songwriting. AMAZON CRACKDOWN: Amazon to require disclosure from publishers who use AI-generated content. SMALL BIZ, BIG GAINS: Small businesses to reap billions in gains from AI, cloud. U.S. President Joe Biden addresses the 76th Session of the U.N. General Assembly on September 21, 2021, at U.N. headquarters in New York City. TRICKY TECH: Biden warns UN to'make sure' AI does not'govern us.' Continue reading… OPEN SEASON: Ex-Google employee launches open-sourced AI protocol to challenge Big Tech.

Hosted by Eleanor Drage and Kerry Mackereth, The Good Robot is a podcast which explores the many complex intersections between gender, feminism and technology. In this episode we talk to Meredith Broussard, data journalism professor at the Arthur L. Carter Institute at New York University. She's also the author of Artificial Unintelligence, which made waves following its release in 2018 by claiming that AI was nothing more than really fancy math. We talk about why we need to bring a little bit more friction back into technology and her latest book More Than a Glitch, which argues that AI that's not designed to be accessible is bad for everyone, in the same way that raised curbs between the pavement and the street that you have to go down to cross the road makes urban outings difficult for lots of people, not just wheelchair users. Data journalist Meredith Broussard is an associate professor at the Arthur L. Carter Journalism Institute of New York University, research director at the NYU Alliance for Public Interest Technology, and the author of several books, including More Than a Glitch: Confronting Race, Gender, and Ability Bias in Tech and Artificial Unintelligence: How Computers Misunderstand the World.

The'America's Got Talent' judge told Fox News Digital why he doesn't like AI technology in songwriting. Amazon will require publishers on Kindle to disclose when any of their content is generated by artificial intelligence after complaints forced the company to take action. "We require you to inform us of AI-generated content (text, images or translations) when you publish a new book or make edits to and republish an existing book through KDP (Kindle Direct Publishing). AI-generated images include cover and interior images and artwork," Amazon said of the updated guidelines, according to a report in Cyber News. The update comes after the company faced complaints from users that some works being sold under the names of human writers contained content that was either fully or partially generated by AI, according to the report.