Annotation and classification of legal text are central components of empirical legal research. Traditionally, these tasks are often delegated to trained research assistants. Motivated by the advances in language modeling, empirical legal scholars are increasingly turning to prompting commercial models, hoping that it will alleviate the significant cost of human annotation. Despite growing use, our understanding of how to best utilize large language models for legal tasks remains limited. We conduct a comprehensive study of 260 legal text classification tasks, nearly all new to the machine learning community. Starting from GPT-4 as a baseline, we show that it has non-trivial but highly varied zero-shot accuracy, often exhibiting performance that may be insufficient for legal work. We then demonstrate that a lightly fine-tuned Llama 3 model vastly outperforms GPT-4 on almost all tasks, typically by double-digit percentage points. We find that larger models respond better to fine-tuning than smaller models. A few tens to hundreds of examples suffice to achieve high classification accuracy. Notably, we can fine-tune a single model on all 260 tasks simultaneously at a small loss in accuracy relative to having a separate model for each task. Our work points to a viable alternative to the predominant practice of prompting commercial models. For concrete legal tasks with some available labeled data, researchers are better off using a fine-tuned open-source model.

Recently, advanced Large Language Models (LLMs) such as GPT-4 have been integrated into many real-world applications like Code Copilot. These applications have significantly expanded the attack surface of LLMs, exposing them to a variety of threats. Among them, jailbreak attacks that induce toxic responses through jailbreak prompts have raised critical safety concerns. To identify these threats, a growing number of red teaming approaches simulate potential adversarial scenarios by crafting jailbreak prompts to test the target LLM. However, existing red teaming methods do not consider the unique vulnerabilities of LLM in different scenarios, making it difficult to adjust the jailbreak prompts to find context-specific vulnerabilities. Meanwhile, these methods are limited to refining jailbreak templates using a few mutation operations, lacking the automation and scalability to adapt to different scenarios. To enable context-aware and efficient red teaming, we abstract and model existing attacks into a coherent concept called "jailbreak strategy" and propose a multi-agent LLM system named RedAgent that leverages these strategies to generate context-aware jailbreak prompts. By self-reflecting on contextual feedback in an additional memory buffer, RedAgent continuously learns how to leverage these strategies to achieve effective jailbreaks in specific contexts. Extensive experiments demonstrate that our system can jailbreak most black-box LLMs in just five queries, improving the efficiency of existing red teaming methods by two times. Additionally, RedAgent can jailbreak customized LLM applications more efficiently. By generating context-aware jailbreak prompts towards applications on GPTs, we discover 60 severe vulnerabilities of these real-world applications with only two queries per vulnerability. We have reported all found issues and communicated with OpenAI and Meta for bug fixes.

Deploying language models (LMs) necessitates outputs to be both high-quality and compliant with safety guidelines. Although Inference-Time Guardrails (ITG) offer solutions that shift model output distributions towards compliance, we find that current methods struggle in balancing safety with helpfulness. ITG Methods that safely address non-compliant queries exhibit lower helpfulness while those that prioritize helpfulness compromise on safety. We refer to this trade-off as the guardrail tax, analogous to the alignment tax. To address this, we propose PrimeGuard, a novel ITG method that utilizes structured control flow. PrimeGuard routes requests to different self-instantiations of the LM with varying instructions, leveraging its inherent instruction-following capabilities and in-context learning. Our tuning-free approach dynamically compiles system-designer guidelines for each query. We construct and release safe-eval, a diverse red-team safety benchmark. Extensive evaluations demonstrate that PrimeGuard, without fine-tuning, overcomes the guardrail tax by (1) significantly increasing resistance to iterative jailbreak attacks and (2) achieving state-of-the-art results in safety guardrailing while (3) matching helpfulness scores of alignment-tuned models. Extensive evaluations demonstrate that PrimeGuard, without fine-tuning, outperforms all competing baselines and overcomes the guardrail tax by improving the fraction of safe responses from 61% to 97% and increasing average helpfulness scores from 4.17 to 4.29 on the largest models, while reducing attack success rate from 100% to 8%. PrimeGuard implementation is available at https://github.com/dynamofl/PrimeGuard and safe-eval dataset is available at https://huggingface.co/datasets/dynamoai/safe_eval.

The people of San Francisco haven't always been kind to Waymo's growing fleet of driverless taxis. The autonomous vehicles, which provide tens of thousands of rides each week, have been torched, stomped on, and verbally berated in recent months. Now Waymo is striking back--in the courts. This month, the Silicon Valley company filed a pair of lawsuits, neither of which have been previously reported, that demand hundreds of thousands of dollars in damages from two alleged vandals. Waymo attorneys said in court papers that the alleged vandalism, which ruined dozens of tires and a tail end, are a significant threat to the company's reputation.

The lawsuits allege Udio produced output with "striking resemblances" to songs including Dancing Queen by ABBA and All I Want For Christmas Is You by Mariah Carey, while Suno allegedly turned out songs similar to I Got You (I Feel Good) by James Brown and Johnny B. Goode by Chuck Berry, among others. Record labels were able to basically recreate versions of very famous songs with highly specific prompts, then linked to them in the lawsuits. I made a short compilation here:https://t.co/9Nu7rW7eqD These lawsuits are not the first to trouble the booming generative AI industry. Visual artists have sued makers of image generating systems, while various newspapers are suing OpenAI, the owner of ChatGPT, for similar allegations.

Apple is failing to effectively monitor its platforms or scan for images and videos of the sexual abuse of children, child safety experts allege, which is raising concerns about how the company can handle growth in the volume of such material associated with artificial intelligence. The UK's National Society for the Prevention of Cruelty to Children (NSPCC) accuses Apple of vastly undercounting how often child sexual abuse material (CSAM) appears in its products. In a year, child predators used Apple's iCloud, iMessage and Facetime to store and exchange CSAM in a higher number of cases in England and Wales alone than the company reported across all other countries combined, according to police data obtained by the NSPCC. Through data gathered via freedom of information requests and shared exclusively with the Guardian, the children's charity found Apple was implicated in 337 recorded offenses of child abuse images between April 2022 and March 2023 in England and Wales. In 2023, Apple made just 267 reports of suspected CSAM on its platforms worldwide to the National Center for Missing & Exploited Children (NCMEC), which is in stark contrast to its big tech peers, with Google reporting more than 1.47m and Meta reporting more than 30.6m, per NCMEC's annual report.

Artificial Intelligence (AI) is one of today's most relevant emergent technologies. In view thereof, this paper proposes that more attention should be paid to the philosophical aspects of AI technology and its use. It is argued that this deficit is generally combined with philosophical misconceptions about the growth of knowledge. To identify these misconceptions, reference is made to the ideas of the philosopher of science Karl Popper and the physicist David Deutsch. The works of both thinkers aim against mistaken theories of knowledge, such as inductivism, empiricism, and instrumentalism. This paper shows that these theories bear similarities to how current AI technology operates. It also shows that these theories are very much alive in the (public) discourse on AI, often called Bayesianism. In line with Popper and Deutsch, it is proposed that all these theories are based on mistaken philosophies of knowledge. This includes an analysis of the implications of these mistaken philosophies for the use of AI in science and society, including some of the likely problem situations that will arise. This paper finally provides a realistic outlook on Artificial General Intelligence (AGI) and three propositions on A(G)I and philosophy (i.e., epistemology).

Large multimodal models (LMMs) are processing increasingly longer and richer inputs. Albeit the progress, few public benchmark is available to measure such development. To mitigate this gap, we introduce LongVideoBench, a question-answering benchmark that features video-language interleaved inputs up to an hour long. Our benchmark includes 3,763 varying-length web-collected videos with their subtitles across diverse themes, designed to comprehensively evaluate LMMs on long-term multimodal understanding. To achieve this, we interpret the primary challenge as to accurately retrieve and reason over detailed multimodal information from long inputs. As such, we formulate a novel video question-answering task termed referring reasoning. Specifically, as part of the question, it contains a referring query that references related video contexts, called referred context. The model is then required to reason over relevant video details from the referred context. Following the paradigm of referring reasoning, we curate 6,678 human-annotated multiple-choice questions in 17 fine-grained categories, establishing one of the most comprehensive benchmarks for long-form video understanding. Evaluations suggest that the LongVideoBench presents significant challenges even for the most advanced proprietary models (e.g. GPT-4o, Gemini-1.5-Pro, GPT-4-Turbo), while their open-source counterparts show an even larger performance gap. In addition, our results indicate that model performance on the benchmark improves only when they are capable of processing more frames, positioning LongVideoBench as a valuable benchmark for evaluating future-generation long-context LMMs.

With the development of large language models (LLMs) like ChatGPT, both their vast applications and potential vulnerabilities have come to the forefront. While developers have integrated multiple safety mechanisms to mitigate their misuse, a risk remains, particularly when models encounter adversarial inputs. This study unveils an attack mechanism that capitalizes on human conversation strategies to extract harmful information from LLMs. We delineate three pivotal strategies: (i) decomposing malicious questions into seemingly innocent sub-questions; (ii) rewriting overtly malicious questions into more covert, benign-sounding ones; (iii) enhancing the harmfulness of responses by prompting models for illustrative examples. Unlike conventional methods that target explicit malicious responses, our approach delves deeper into the nature of the information provided in responses. Through our experiments conducted on GPT-3.5-turbo, GPT-4, and Llama2, our method has demonstrated a marked efficacy compared to conventional attack methods. In summary, this work introduces a novel attack method that outperforms previous approaches, raising an important question: How to discern whether the ultimate intent in a dialogue is malicious?

One thing seems for sure: people are not ready for "digital workers" just yet. That's the lesson learned by Sarah Franklin, the CEO of Lattice, a human resources and performance management platform that offers performance coaching, talent reviews, onboarding automation, compensation management and a host of other HR tools to more than 5,000 organizations around the world. What is a digital employee? According to Franklin, it's avatars like Devin the engineer, Harvey the lawyer, Einstein the service agent and Piper the sales agent who have "entered the workforce and become our colleagues". But these are not real workers.