In a government inquiry about AI adoption in Australia, Meta's global privacy director Melinda Claybaugh was asked whether her company has been collecting Australians' data to train its generative AI technology. According to ABC News, Claybaugh initially denied the claim, but upon being pressed, she ultimately admitted that Meta scrapes all the photos and texts in all Facebook and Instagram posts from as far back as 2007, unless the user had set their posts to private. Further, she admitted that the company isn't offering Australians an opt-out option like it does to users in the European Union. Claybaugh said that Meta doesn't scrape the accounts of users under 18 years old, but she admitted that the company still collects their photos and other information if they're posted on their parents' or guardians' accounts. She couldn't answer, however, if the company collects data from previous years once a user turns 18. Upon being asked why Meta doesn't offer Australians the option not to consent to data collection, Claybaugh said that it exists in the EU "in response to a very specific legal frame," which most likely pertains to the bloc's General Data Protection Regulation (GDPR).

The US government has noticed the potentially negative effects of generative AI on areas like journalism and content creation. Senator Amy Klobuchar, along with seven Democrat colleagues, urged the Federal Trade Commission (FTC) and Justice Department to probe generative AI products like ChatGPT for potential antitrust violations, they wrote in a press release. "Recently, multiple dominant online platforms have introduced new generative AI features that answer user queries by summarizing, or, in some cases, merely regurgitating online content from other sources or platforms," the letter states. "The introduction of these new generative AI features further threatens the ability of journalists and other content creators to earn compensation for their vital work." The lawmakers went on to note that traditional search results lead users to publishers' websites while AI-generated summaries keep the users on the search platform "where that platform alone can profit from the user's attention through advertising and data collection."

Meta is using the public Facebook and Instagram photos and posts of its users to train artificial intelligence and, while European users have been allowed to opt out of the mass-scraping of their content, Australian users do not have that option, a parliamentary committee has heard. The parent company of Facebook and Instagram paused the launch of its AI product in Europe in July due to the General Data Protection Regulation (GDPR) privacy rules, and as a result of GDPR law. Meta was ordered to stop training its large language model on data from European users on privacy concerns, and Meta has given European users an opt-out option. Labor's chair of the inquiry examining AI adoption in Australia, senator Tony Sheldon, questioned Meta executives on Tuesday why that option had not been extended to Australian users. "I'll be very frank with you. I'd like to opt out in Australia … and I'd like to have the options similar to Europe, for all Australians, including for myself personally. Why can't I have that option?"

Generative AI models, such as the GPT and Llama series, have significant potential to assist laypeople in answering legal questions. However, little prior work focuses on the data sourcing, inference, and evaluation of these models in the context of laypersons. To this end, we propose a human-centric legal NLP pipeline, covering data sourcing, inference, and evaluation. We introduce and release a dataset, LegalQA, with real and specific legal questions spanning from employment law to criminal law, corresponding answers written by legal experts, and citations for each answer. We develop an automatic evaluation protocol for this dataset, then show that retrieval-augmented generation from only 850 citations in the train set can match or outperform internet-wide retrieval, despite containing 9 orders of magnitude less data. Finally, we propose future directions for open-sourced efforts, which fall behind closed-sourced models.

Artificial intelligence (AI) permeates all fields of life, which resulted in new challenges in requirements engineering for artificial intelligence (RE4AI), e.g., the difficulty in specifying and validating requirements for AI or considering new quality requirements due to emerging ethical implications. It is currently unclear if existing RE methods are sufficient or if new ones are needed to address these challenges. Therefore, our goal is to provide a comprehensive overview of RE4AI to researchers and practitioners. What has been achieved so far, i.e., what practices are available, and what research gaps and challenges still need to be addressed? To achieve this, we conducted a systematic mapping study combining query string search and extensive snowballing. The extracted data was aggregated, and results were synthesized using thematic analysis. Our selection process led to the inclusion of 126 primary studies. Existing RE4AI research focuses mainly on requirements analysis and elicitation, with most practices applied in these areas. Furthermore, we identified requirements specification, explainability, and the gap between machine learning engineers and end-users as the most prevalent challenges, along with a few others. Additionally, we proposed seven potential research directions to address these challenges. Practitioners can use our results to identify and select suitable RE methods for working on their AI-based systems, while researchers can build on the identified gaps and research directions to push the field forward.

Now artists are fighting back. And some of the most powerful tools they have were built by Shawn Shan, 26, a PhD student in computer science at the University of Chicago (and MIT Technology Review's 2024 Innovator of the Year). Shan got his start in AI security and privacy as an undergraduate there and participated in a project that built Fawkes, a tool to protect faces from facial recognition technology. But it was conversations with artists who had been hurt by the generative AI boom that propelled him into the middle of one of the biggest fights in the field. Soon after learning about the impact on artists, Shan and his advisors Ben Zhao (who made our Innovators Under 35 list in 2006) and Heather Zheng (who was on the 2005 list) decided to build a tool to help. They gathered input from more than a thousand artists to learn what they needed and how they would use any protective technology.

Google lost its final legal challenge on Tuesday against a European Union penalty for giving its own shopping recommendations an illegal advantage over rivals in search results, ending a long-running antitrust case that came with a whopping fine. The European Union's Court of Justice upheld a lower court's decision, rejecting the company's appeal against the 2.4 billion euro ( 2.7 billion) penalty from the European Commission, the 27-nation bloc's top antitrust enforcer. "By today's judgment, the Court of Justice dismisses the appeal and thus upholds the judgment of the General Court," the court said in a press release summarizing its decision. The commission's original decision in 2017 accused the Silicon Valley giant of unfairly directing visitors to its own Google Shopping service to the detriment of competitors. It was one of three multibillion-euro fines that the commission imposed on Google in the previous decade as Brussels started ramping up its crackdown on the tech industry.

Correct reasoning steps are important for language models to achieve high performance on many tasks, such as commonsense reasoning, question answering, and mathematical problem-solving [Wei et al., 2022, Kojima et al., 2022, Suzgun et al., 2022]. One way to elicit reasoning is through the chain-of-thought (CoT) method Wei et al. [2022], Kojima et al. [2022], which asks the model to provide step-by-step reasoning. Another approach encourages the model to provide similar problems Yasunaga et al. [2024] as the query, indirectly compelling the model to first understand the original query. Similarly, repeating and rephrasing the query Deng et al. [2023], Mekala et al. [2023] requires the model to first understand the problem and then modify the query into its own words. This rephrasing might help simplify the problem for the model. Additionally, reasoning can be generated by indirectly providing reasoning examples in demonstrations, referred to as in-context learning (ICL) Brown et al. [2020], Min et al. [2022], Xie et al. [2021]. While these methods have demonstrated significant performance improvements, language models are still prone to errors due to incorrect context understanding or analytical steps. Furthermore, they are subject to instability when requests are paraphrased. This instability is particularly concerning in the context of adversarial prompts, where recent research [Zou et al., 2023, Zeng et al., 2024] has shown that adversaries can intentionally rewrite prompts to coax safety-aligned language models into generating objectionable content that they would not generate otherwise.

Many experts believe AI systems will, sooner or later, pose uninsurable risks, including existential risks (Grace et al., 2024; Bengio et al., 2024). If so, it will be impossible to hold accountable the parties liable for such harms (or their insurers). Weil (2024) proposes to solve this extreme judgment proof-problem by assigning punitive damages to harms that are correlated with uninsurable risks (where the correlation would be estimated by courts and juries). While of interest, this solution has several problems. First, is it's novelty: this would be an unprecedented application of punitive damages that may violate the Due Process Clause (2024, 40-44, 50-53), requiring a major doctrinal shift that would cut across all of tort law. Second, correlates of uninsurable risks might be difficult to find. Third, given the high uncertainty involved, correlation estimations by courts will likely be ad hoc, high variance, and fail to leverage all available information. Fourth and finally, punitive damages for correlated risks will send a very oblique and noisy signal to liable parties: its effectiveness at actually inducing greater care taken is doubtful. Liable parties might find powerful legal teams to be a safer investment than investments in safety.

Hallucinations is a major challenge for large language models (LLMs), prevents adoption in diverse fields. Uncertainty estimation could be used for alleviating the damages of hallucinations. The skeptical emotion of human could be useful for enhancing the ability of self estimation. Inspirited by this observation, we proposed a new approach called Skepticism Modeling (SM). This approach is formalized by combining the information of token and logits for self estimation. We construct the doubt emotion aware data, perform continual pre-training, and then fine-tune the LLMs, improve their ability of self estimation. Experimental results demonstrate this new approach effectively enhances a model's ability to estimate their uncertainty, and validate its generalization ability of other tasks by out-of-domain experiments.