The strong performance (Chowdhery et al., 2022; Nostalgebraist, 2022; OpenAI, 2023; Google, 2023), and emergent abilities (Wei et al., 2022) of modern language models (LMs) depend on self-supervised pretraining on massive text datasets. All model developers implicitly or explicitly decide the composition of these datasets: what data sources to include, whether to filter for attributes such as quality and toxicity, and when to gather new documents. While many of the most prominent models do not document their curation procedures (OpenAI, 2023; Google, 2023), or only document which procedures they used (Brown et al., 2020; Nostalgebraist, 2022; Scao et al., 2022; Touvron et al., 2023), they rarely document why they chose those protocols or what effect they had. This documentation debt leaves practitioners to be guided by intuitions and precedents, neither thoroughly evaluated (Bandy and Vincent, 2021; Sambasivan et al., 2021). Given the outsized and fundamental role of pretraining data in modern LMs, we believe this neglectful practice has detracted from responsible data use and hampered effective model development (Rogers, 2021; Gebru et al., 2021; Bender and Friedman, 2018). Among the small number of general-purpose LMs dominating community use and discussion, the prevailing focus has been on the scale of pretraining data and number of optimization steps (Brown et al., 2020; Nostalgebraist, 2022; Google, 2023). In this work, we systematically test how common data design decisions affect model performance--specifically: the time of collection, content filtering strategy (toxicity/quality), and domain composition. We study the impacts in two ways.

The rise of generative artificial intelligence (AI) has sparked concerns about its potential influence on unemployment and market depression. This study addresses this concern by examining the impact of generative AI on product markets. To overcome the challenge of causal inference, given the inherent limitations of conducting controlled experiments, this paper identifies an unanticipated and sudden leak of a highly proficient image-generative AI as a novel instance of a "natural experiment". This AI leak spread rapidly, significantly reducing the cost of generating anime-style images compared to other styles, creating an opportunity for comparative assessment. We collect real-world data from an artwork outsourcing platform. Surprisingly, our results show that while generative AI lowers average prices, it substantially boosts order volume and overall revenue. This counterintuitive finding suggests that generative AI confers benefits upon artists rather than detriments. The study further offers theoretical economic explanations to elucidate this unexpected phenomenon. By furnishing empirical evidence, this paper dispels the notion that generative AI might engender depression, instead underscoring its potential to foster market prosperity. These findings carry significant implications for practitioners, policymakers, and the broader AI community.

Recently, diffusion-based deep generative models (e.g., Stable Diffusion) have shown impressive results in text-to-image synthesis. However, current text-to-image models often require multiple passes of prompt engineering by humans in order to produce satisfactory results for real-world applications. We propose BeautifulPrompt, a deep generative model to produce high-quality prompts from very simple raw descriptions, which enables diffusion-based models to generate more beautiful images. In our work, we first fine-tuned the BeautifulPrompt model over low-quality and high-quality collecting prompt pairs. Then, to ensure that our generated prompts can generate more beautiful images, we further propose a Reinforcement Learning with Visual AI Feedback technique to fine-tune our model to maximize the reward values of the generated prompts, where the reward values are calculated based on the PickScore and the Aesthetic Scores. Our results demonstrate that learning from visual AI feedback promises the potential to improve the quality of generated prompts and images significantly. We further showcase the integration of BeautifulPrompt to a cloud-native AI platform to provide better text-to-image generation service in the cloud.

Drones, hidden cameras, thermal vision scopes--these are just a few examples of the high-tech equipment recommended by the animal liberation group Direct Action Everywhere, according to a manual released by the organization this week. The document, which was reviewed by WIRED, is a rare glimpse into how the organization is using tech to target factory farms in often brazen operations that have rescued pigs, goats, ducks, and chickens. Extremist groups are experimenting with generative AI to flood social media with propaganda and misinformation, researchers at Tech Against Terrorism have told WIRED. A new report from the group details how, in recent months, terrorists and other extremist organizations have been using artificial intelligence to manipulate imagery and thwart content moderation. As platforms have struggled to keep up with this flood of extremist content, a new tool called Altitude, built in collaboration between Tech Against Terrorism and Google, is seeking to address the problem.

Creating and disseminating convincing propaganda used to require the resources of a state. Now all it takes is a smartphone. Generative artificial intelligence is now capable of creating fake pictures, clones of our voices, and even videos depicting and distorting world events. The result: From our personal circles to the political circuses, everyone must now question whether what they see and hear is true.

Microsoft temporarily prohibited its employees from using ChatGPT "due to security and data concerns," according to CNBC. The company announced the rule in an internal website and even blocked corporate devices from being able to access the AI chatbot. While several tech companies had prohibited -- or had at least discouraged -- the internal use of ChatGPT in the past, Microsoft doing the same thing was certainly curious, seeing as it's OpenAI's biggest and most prominent investor. In January, Microsoft pledged to invest $10 billion in ChatGPT's developer over the next few years after pouring $3 billion into the company in the past. The AI-powered tools it rolled out for its products, such as Bing's chatbot, also use OpenAI's large language model.

The creator of a viral video purporting to show Japanese Prime Minister Fumio Kishida making explicit sexual admissions in a live news broadcast said he made it in about an hour as a "little joke." "I didn't think it would create such a stir," the man in his 20s said of the video, explaining that he used generative artificial intelligence technology to create Kishida's voice and mouth movements. The video shows the prime minister speaking to the camera during a live news program on Japanese broadcaster Nippon Television Network. The company's logo appears in the top right corner of the screen along with a ticker saying, "Breaking News."

Humane, the startup founded by former Apple design and engineering team Imran Chaudhri and Bethany Bongiorno, has officially launched its long-awaited Ai Pin -- making a splashy foray into the nascent field of artificial intelligence hardware. The device can magnetically clip onto clothing and will cost $699 with a $24-a-month subscription -- which will come with unlimited data and phone calls. The company also said it would partner with T-Mobile for phone service and Microsoft and OpenAI for AI technology. The device will be available to order starting Nov. 16. "For the technology you are getting, we set a high bar for ourselves in terms of pricing it at a level we think is approachable and accessible," Bongiorno, Humane's chief executive officer, said in an interview on Bloomberg TV Thursday.

Text-to-image generative models such as Stable Diffusion and DALL$\cdot$E raise many ethical concerns due to the generation of harmful images such as Not-Safe-for-Work (NSFW) ones. To address these ethical concerns, safety filters are often adopted to prevent the generation of NSFW images. In this work, we propose SneakyPrompt, the first automated attack framework, to jailbreak text-to-image generative models such that they generate NSFW images even if safety filters are adopted. Given a prompt that is blocked by a safety filter, SneakyPrompt repeatedly queries the text-to-image generative model and strategically perturbs tokens in the prompt based on the query results to bypass the safety filter. Specifically, SneakyPrompt utilizes reinforcement learning to guide the perturbation of tokens. Our evaluation shows that SneakyPrompt successfully jailbreaks DALL$\cdot$E 2 with closed-box safety filters to generate NSFW images. Moreover, we also deploy several state-of-the-art, open-source safety filters on a Stable Diffusion model. Our evaluation shows that SneakyPrompt not only successfully generates NSFW images, but also outperforms existing text adversarial attacks when extended to jailbreak text-to-image generative models, in terms of both the number of queries and qualities of the generated NSFW images. SneakyPrompt is open-source and available at this repository: \url{https://github.com/Yuchen413/text2image_safety}.

Extremist groups have begun to experiment with artificial intelligence, and in particular generative AI, in order to create a flood of new propaganda. Experts now fear the growing use of generative AI tools by these groups will overturn the work Big Tech has done in recent years to keep their content off the internet. "Our biggest concern is that if terrorists start using gen AI to manipulate imagery at scale, this could well destroy hash-sharing as a solution," Adam Hadley, the executive director of Tech Against Terrorism, tells WIRED. "This is a massive risk." For years, Big Tech platforms have worked hard to create databases of known violent extremist content, known as hashing databases, which are shared across platforms to quickly and automatically remove such content from the internet.