Aligning large language models (LLMs) with human expectations requires high-quality instructional dialogues, which can be achieved by raising diverse, in-depth, and insightful instructions that deepen interactions. Existing methods target instructions from real instruction dialogues as a learning goal and fine-tune a user simulator for posing instructions. However, the user simulator struggles to implicitly model complex dialogue flows and pose high-quality instructions. In this paper, we take inspiration from the cognitive abilities inherent in human learning and propose the explicit modeling of complex dialogue flows through instructional strategy reuse. Specifically, we first induce high-level strategies from various real instruction dialogues. These strategies are applied to new dialogue scenarios deductively, where the instructional strategies facilitate high-quality instructions. Experimental results show that our method can generate diverse, in-depth, and insightful instructions for a given dialogue history. The constructed multi-turn instructional dialogues can outperform competitive baselines on the downstream chat model.

There's a moment in the new hit Amazon Prime Video series Fallout where the sunny protagonist, having emerged from her underground commune into a postapocalyptic hellscape, tries to convince a bloodthirsty mutant to follow the Golden Rule, to do unto others as you would have them do unto you. I expected the mutant--or Ghoul, to be more precise--to shoot back some nihilistic platitude in return, maybe a slang-ified version of a Thomas Hobbes quote. Instead, we get a perfect line: "Yeah, well, the wasteland's got its own golden rule," he replies. "'Thou shalt get sidetracked by bullshit every goddamn time.'" That rejoinder distills what makes Fallout, both the video game series and its television adaptation, so great. After all, getting sidetracked by bullshit is what Fallout has always been about.

Amazon Music is joining Spotify in starting to offer a generative AI-powered playlist feature. For now, Maestro is available in beta to a small number of Amazon Music users in the US on iOS and Android. Folks who are included in the beta will see Maestro on the home screen after they update to the latest version of the app. They can also access the tool by tapping the plus button to create a new playlist. The idea is to use natural language prompts to create any kind of playlist imaginable.

Fox News correspondent Stephanie Bennett joins'Fox News Live' to break down recent evidence tying missile fragments in Russian attacks to North Korea. North Korea is putting surveillance cameras in schools and workplaces and collecting fingerprints, photographs and other biometric information from its citizens in a technology-driven push to monitor its population even more closely, a report said Tuesday. The state's growing use of digital surveillance tools, which combine equipment imported from China with domestically developed software, threatens to erase many of the small spaces North Koreans have left to engage in private business activities, access foreign media and secretly criticize their government, the researchers wrote. But the isolated country's digital ambitions have to contend with poor electricity supplies and low network connectivity. Those challenges, and a history of reliance on human methods of spying on its citizens, mean that digital surveillance isn't yet as pervasive as in China, according to the report, published by the North Korea-focused website 38 North. The study's findings align with widely held views that North Korean leader Kim Jong Un is stepping up efforts to tighten the state's control of its citizens and promote loyalty to his regime.

An illustration of a large envelope fills the screen--I've received a new e-mail, my first in weeks. I click on the middle of the envelope and a note opens in size thirty-six font. It's a top-secret assignment for me, a renegade ex-C.I.A. agent who can kick higher than anyone else in the agency. "Looks like this old dog is heading back to the pound," I growl, and close the e-mail by turning off my entire computer. I pull up a digitized photo on the screen.

Netflix has been accused of using AI-manipulated imagery in the true crime documentary What Jennifer Did, Futurism has reported. Several photos show typical signs of AI trickery, including mangled hands, strange artifacts and more. If accurate, the report raises serious questions about the use of such images in documentaries, particularly since the person depicted is currently in prison awaiting retrial. In one egregious image, the left hand of the documentary's subject Jennifer Pan is particularly mangled, while another image shows a strange gap in her cheek. Netflix has yet to acknowledge the report, but the images show clear signs of manipulation and were never labeled as AI-generated. The AI may be generating the imagery based on real photos of Pan, as PetaPixel suggested.

The era of pre-trained models has ushered in a wealth of new insights for the machine learning community. Among the myriad of questions that arise, one of paramount importance is: 'Do pre-trained models possess comprehensive knowledge?' This paper seeks to address this crucial inquiry. In line with our objective, we have made publicly available a novel dataset comprised of images from TV series released post-2021. This dataset holds significant potential for use in various research areas, including the evaluation of incremental learning, novel class discovery, and long-tailed learning, among others. Project page: https://tv-100.github.io/

The open-sourcing of large language models (LLMs) accelerates application development, innovation, and scientific progress. This includes both base models, which are pre-trained on extensive datasets without alignment, and aligned models, deliberately designed to align with ethical standards and human values. Contrary to the prevalent assumption that the inherent instruction-following limitations of base LLMs serve as a safeguard against misuse, our investigation exposes a critical oversight in this belief. By deploying carefully designed demonstrations, our research demonstrates that base LLMs could effectively interpret and execute malicious instructions. To systematically assess these risks, we introduce a novel set of risk evaluation metrics. Empirical results reveal that the outputs from base LLMs can exhibit risk levels on par with those of models fine-tuned for malicious purposes. This vulnerability, requiring neither specialized knowledge nor training, can be manipulated by almost anyone, highlighting the substantial risk and the critical need for immediate attention to the base LLMs' security protocols.

The quadratic complexity and weak length extrapolation of Transformers limits their ability to scale to long sequences, and while sub-quadratic solutions like linear attention and state space models exist, they empirically underperform Transformers in pretraining efficiency and downstream task accuracy. We introduce Megalodon, a neural architecture for efficient sequence modeling with unlimited context length. Megalodon inherits the architecture of Mega (exponential moving average with gated attention), and further introduces multiple technical components to improve its capability and stability, including complex exponential moving average (CEMA), timestep normalization layer, normalized attention mechanism and pre-norm with two-hop residual configuration. In a controlled head-to-head comparison with Llama2, Megalodon achieves better efficiency than Transformer in the scale of 7 billion parameters and 2 trillion training tokens. Megalodon reaches a training loss of 1.70, landing mid-way between Llama2-7B (1.75) and 13B (1.67). Code: https://github.com/XuezheMax/megalodon

As generative AI progresses rapidly, new synthetic image generators continue to emerge at a swift pace. Traditional detection methods face two main challenges in adapting to these generators: the forensic traces of synthetic images from new techniques can vastly differ from those learned during training, and access to data for these new generators is often limited. To address these issues, we introduce the Ensemble of Expert Embedders (E3), a novel continual learning framework for updating synthetic image detectors. E3 enables the accurate detection of images from newly emerged generators using minimal training data. Our approach does this by first employing transfer learning to develop a suite of expert embedders, each specializing in the forensic traces of a specific generator. Then, all embeddings are jointly analyzed by an Expert Knowledge Fusion Network to produce accurate and reliable detection decisions. Our experiments demonstrate that E3 outperforms existing continual learning methods, including those developed specifically for synthetic image detection.