A California jury has found that Tesla was not at fault for a fatal 2019 crash that allegedly involved its Autopilot system, in the first US trial yet for a case claiming its software directly caused a death. The lawsuit alleged Tesla knowingly shipped out cars with a defective Autopilot system, leading to a crash that killed a Model 3 owner and severely injured two passengers, Reuters reports. Per the lawsuit, 37-year-old Micah Lee was driving his Tesla Model 3 on a highway outside of Los Angeles at 65 miles per hour when it turned sharply off the road and slammed into a palm tree before catching fire. Lee died in the crash. The company was sued for $400 million plus punitive damages by Lee's estate and the two surviving victims, including a boy who was 8 years old at the time and was disemboweled in the accident, according to an earlier report from Reuters.

Tesla on Tuesday won the first US trial over allegations that its autopilot driver assistance feature led to a death, a major victory for the automaker as it faces several similar lawsuits across the country. The case, in a California state court, was filed by two passengers in a 2019 crash who accused the company of knowing the autopilot feature was defective when it sold the car. Tesla argued that human error caused the crash. The 12-member jury on Tuesday announced they had found the vehicle did not have a manufacturing defect. The verdict came on the fourth day of deliberations, and the vote was 9-3.

This week, UK prime minister Rishi Sunak is hosting a group of 100 representatives from the worlds of business and politics to discuss the potential and pitfalls of artificial intelligence. The AI Safety Summit, held at Bletchley Park, UK, begins on 1 November and aims to come up with a set of global principles with which to develop and deploy "frontier AI models" – the terminology favoured by Sunak and key figures in the AI industry for powerful models that don't yet exist, but may be built very soon. While the Bletchley Park event is the focal point, there is a wider week of fringe events being held in the UK, alongside a raft of UK government announcements on AI. Here are the latest developments. The global community has decided that the week of the UK summit is a ripe time to announce their own AI developments.

Kurt "The Cyberguy" Knutsson explains how facial recognition technology can help you find your perfect match. FIRST ON FOX: House lawmakers are eyeing legislation to make dating app users more aware of potential scammers who tricked victims out of more than $1 billion in a single year. Rep. David Valadao, R-Calif., is reintroducing his Online Dating Safety Act this week alongside Rep. Brittany Pettersen, D-Colo. If passed the bill would force dating apps and services to send users a fraud notification when they have interacted with someone banned from the app for using a fake identity or using the app to defraud others. Dating apps seen on the screen of an iPhone.

In his speech Thursday, Sunak announced a new global AI Safety Institute in Britain that would "carefully examine, evaluate, and test new types of AI." He offered few specifics about how the agency would function, and whether there would be any legal requirements for companies to submit their models for assessments. Sunak said the British government has invested 1 billion pounds in supercomputing, 2.5 billion pounds in quantum computers, and he announced an investment of 100 million pounds into the use of AI to discover treatments for diseases.

As AI takes on a greater role in the modern world, it is essential to ensure that AI models can overcome decision uncertainty and remain aligned with human morality and interests. This research paper proposes a method for improving the decision-making of language models (LMs) via Automated Parliaments (APs) - constructs made of AI delegates each representing a certain perspective. Delegates themselves consist of three AI models: generators, modifiers, and evaluators. We specify two mechanisms for producing optimal solutions: the Simultaneous Modification mechanism for response creation and an evaluation mechanism for fairly assessing solutions. The overall process begins when each generator creates a response aligned with its delegate's theory. The modifiers alter all other responses to make them more self-aligned. The evaluators collectively assess the best end response. Finally, the modifiers and generators learn from feedback from the evaluators. In our research, we tested the evaluation mechanism, comparing the use of single-value zero-shot prompting and AP few-shot prompting in evaluating morally contentious scenarios. We found that the AP architecture saw a 57.3% reduction in its loss value compared to the baseline. We conclude by discussing some potential applications of APs and specifically their potential impact when implemented as Automated Moral Parliaments.

Releasing court decisions to the public relies on proper anonymization to protect all involved parties, where necessary. The Swiss Federal Supreme Court relies on an existing system that combines different traditional computational methods with human experts. In this work, we enhance the existing anonymization software using a large dataset annotated with entities to be anonymized. We compared BERT-based models with models pre-trained on in-domain data. Our results show that using in-domain data to pre-train the models further improves the F1-score by more than 5\% compared to existing models. Our work demonstrates that combining existing anonymization methods, such as regular expressions, with machine learning can further reduce manual labor and enhance automatic suggestions.

AI developers often apply safety alignment procedures to prevent the misuse of their AI systems. For example, before Meta released Llama 2-Chat, a collection of instruction fine-tuned large language models, they invested heavily in safety training, incorporating extensive red-teaming and reinforcement learning from human feedback. However, it remains unclear how well safety training guards against model misuse when attackers have access to model weights. We explore the robustness of safety training in language models by subversively fine-tuning the public weights of Llama 2-Chat. We employ low-rank adaptation (LoRA) as an efficient fine-tuning method. With a budget of less than $200 per model and using only one GPU, we successfully undo the safety training of Llama 2-Chat models of sizes 7B, 13B, and 70B. Specifically, our fine-tuning technique significantly reduces the rate at which the model refuses to follow harmful instructions. We achieve a refusal rate below 1% for our 70B Llama 2-Chat model on two refusal benchmarks. Our fine-tuning method retains general performance, which we validate by comparing our fine-tuned models against Llama 2-Chat across two benchmarks. Additionally, we present a selection of harmful outputs produced by our models. While there is considerable uncertainty about the scope of risks from current models, it is likely that future models will have significantly more dangerous capabilities, including the ability to hack into critical infrastructure, create dangerous bio-weapons, or autonomously replicate and adapt to new environments. We show that subversive fine-tuning is practical and effective, and hence argue that evaluating risks from fine-tuning should be a core part of risk assessments for releasing model weights.

Recent technological advancements have led to a large number of patents in a diverse range of domains, making it challenging for human experts to analyze and manage. State-of-the-art methods for multi-label patent classification rely on deep neural networks (DNNs), which are complex and often considered black-boxes due to their opaque decision-making processes. In this paper, we propose a novel deep explainable patent classification framework by introducing layer-wise relevance propagation (LRP) to provide human-understandable explanations for predictions. We train several DNN models, including Bi-LSTM, CNN, and CNN-BiLSTM, and propagate the predictions backward from the output layer up to the input layer of the model to identify the relevance of words for individual predictions. Considering the relevance score, we then generate explanations by visualizing relevant words for the predicted patent class. Experimental results on two datasets comprising two-million patent texts demonstrate high performance in terms of various evaluation measures. The explanations generated for each prediction highlight important relevant words that align with the predicted class, making the prediction more understandable. Explainable systems have the potential to facilitate the adoption of complex AI-enabled methods for patent classification in real-world applications.

Embedding covert streams into a cover channel is a common approach to circumventing Internet censorship, due to censors' inability to examine encrypted information in otherwise permitted protocols (Skype, HTTPS, etc.). However, recent advances in machine learning (ML) enable detecting a range of anti-censorship systems by learning distinct statistical patterns hidden in traffic flows. Therefore, designing obfuscation solutions able to generate traffic that is statistically similar to innocuous network activity, in order to deceive ML-based classifiers at line speed, is difficult. In this paper, we formulate a practical adversarial attack strategy against flow classifiers as a method for circumventing censorship. Specifically, we cast the problem of finding adversarial flows that will be misclassified as a sequence generation task, which we solve with Amoeba, a novel reinforcement learning algorithm that we design. Amoeba works by interacting with censoring classifiers without any knowledge of their model structure, but by crafting packets and observing the classifiers' decisions, in order to guide the sequence generation process. Our experiments using data collected from two popular anti-censorship systems demonstrate that Amoeba can effectively shape adversarial flows that have on average 94% attack success rate against a range of ML algorithms. In addition, we show that these adversarial flows are robust in different network environments and possess transferability across various ML models, meaning that once trained against one, our agent can subvert other censoring classifiers without retraining.