The social media platform, formerly Twitter, was so slow to react that one image racked up 47m views before it was taken down. It was largely Swift's fans who mobilised and mass-reported the images, and there was a sense of public anger, with even the White House calling it "alarming". X eventually removed the images and blocked searches to the pop star's name on Sunday evening. For women who have been victims of the creation and sharing of nonconsensual deepfake pornography, the events of the past week will have been a horrible reminder of their own abuse, even if they may also hope that the spotlight will force legislators into action. But because the pictures were removed, Swift's experience is far from the norm.

A bipartisan group of US senators introduced a bill Tuesday that would criminalize the spread of nonconsensual, sexualized images generated by artificial intelligence. The measure comes in direct response to the proliferation of pornographic AI-made images of Taylor Swift on X, formerly Twitter, in recent days. The measure would allow victims depicted in nude or sexually explicit "digital forgeries" to seek a civil penalty against "individuals who produced or possessed the forgery with intent to distribute it" or anyone who received the material knowing it was not made with consent. Dick Durbin, the US Senate majority whip, and senators Lindsey Graham, Amy Klobuchar and Josh Hawley are behind the bill, known as the Disrupt Explicit Forged Images and Non-Consensual Edits Act of 2024, or the "Defiance Act." "This month, fake, sexually-explicit images of Taylor Swift that were generated by artificial intelligence swept across social media platforms. Although the imagery may be fake, the harm to the victims from the distribution of sexually-explicit'deepfakes' is very real," Durbin said in a press release.

Language models have become a critical technology to tackling a wide range of natural language processing tasks, yet many details about how the best-performing language models were developed are not reported. In particular, information about their pretraining corpora is seldom discussed: commercial language models rarely provide any information about their data; even open models rarely release datasets they are trained on, or an exact recipe to reproduce them. As a result, it is challenging to conduct certain threads of language modeling research, such as understanding how training data impacts model capabilities and shapes their limitations. To facilitate open research on language model pretraining, we release Dolma, a three trillion tokens English corpus, built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. In addition, we open source our data curation toolkit to enable further experimentation and reproduction of our work. In this report, we document Dolma, including its design principles, details about its construction, and a summary of its contents. We interleave this report with analyses and experimental results from training language models on intermediate states of Dolma to share what we have learned about important data curation practices, including the role of content or quality filters, deduplication, and multi-source mixing. Dolma has been used to train OLMo, a state-of-the-art, open language model and framework designed to build and study the science of language modeling.

Recent work in Human-Robot Interaction (HRI) has shown that robots can leverage implicit communicative signals from users to understand how they are being perceived during interactions. For example, these signals can be gaze patterns, facial expressions, or body motions that reflect internal human states. To facilitate future research in this direction, we contribute the REACT database, a collection of two datasets of human-robot interactions that display users' natural reactions to robots during a collaborative game and a photography scenario. Further, we analyze the datasets to show that interaction history is an important factor that can influence human reactions to robots. As a result, we believe that future models for interpreting implicit feedback in HRI should explicitly account for this history. REACT opens up doors to this possibility in the future.

Training task-oriented dialog agents based on reinforcement learning is time-consuming and requires a large number of interactions with real users. How to grasp dialog policy within limited dialog experiences remains an obstacle that makes the agent training process less efficient. In addition, most previous frameworks start training by randomly choosing training samples, which differs from the human learning method and hurts the efficiency and stability of training. Therefore, we propose Scheduled Curiosity-Deep Dyna-Q (SC-DDQ), a curiosity-driven curriculum learning framework based on a state-of-the-art model-based reinforcement learning dialog model, Deep Dyna-Q (DDQ). Furthermore, we designed learning schedules for SC-DDQ and DDQ, respectively, following two opposite training strategies: classic curriculum learning and its reverse version. Our results show that by introducing scheduled learning and curiosity, the new framework leads to a significant improvement over the DDQ and Deep Q-learning(DQN). Surprisingly, we found that traditional curriculum learning was not always effective. Specifically, according to the experimental results, the easy-first and difficult-first strategies are more suitable for SC-DDQ and DDQ. To analyze our results, we adopted the entropy of sampled actions to depict action exploration and found that training strategies with high entropy in the first stage and low entropy in the last stage lead to better performance.

Anticipating the negative impacts of emerging AI technologies is a challenge, especially in the early stages of development. An understudied approach to such anticipation is the use of LLMs to enhance and guide this process. Despite advancements in LLMs and evaluation metrics to account for biases in generated text, it is unclear how well these models perform in anticipatory tasks. Specifically, the use of LLMs to anticipate AI impacts raises questions about the quality and range of categories of negative impacts these models are capable of generating. In this paper we leverage news media, a diverse data source that is rich with normative assessments of emerging technologies, to formulate a taxonomy of impacts to act as a baseline for comparing against. By computationally analyzing thousands of news articles published by hundreds of online news domains around the world, we develop a taxonomy consisting of ten categories of AI impacts. We then evaluate both instruction-based (GPT-4 and Mistral-7B-Instruct) and fine-tuned completion models (Mistral-7B and GPT-3) using a sample from this baseline. We find that the generated impacts using Mistral-7B, fine-tuned on impacts from the news media, tend to be qualitatively on par with impacts generated using a larger scale model such as GPT-4. Moreover, we find that these LLMs generate impacts that largely reflect the taxonomy of negative impacts identified in the news media, however the impacts produced by instruction-based models had gaps in the production of certain categories of impacts in comparison to fine-tuned models. This research highlights a potential bias in state-of-the-art LLMs when used for anticipating impacts and demonstrates the advantages of aligning smaller LLMs with a diverse range of impacts, such as those reflected in the news media, to better reflect such impacts during anticipatory exercises.

The increasing reliance on AI-driven solutions, particularly Large Language Models (LLMs) like the GPT series, for information retrieval highlights the critical need for their factuality and fairness, especially amidst the rampant spread of misinformation and disinformation online. Our study evaluates the factual accuracy, stability, and biases in widely adopted GPT models, including GPT-3.5 and GPT-4, contributing to reliability and integrity of AI-mediated information dissemination. We introduce 'Global-Liar,' a dataset uniquely balanced in terms of geographic and temporal representation, facilitating a more nuanced evaluation of LLM biases. Our analysis reveals that newer iterations of GPT models do not always equate to improved performance. Notably, the GPT-4 version from March demonstrates higher factual accuracy than its subsequent June release. Furthermore, a concerning bias is observed, privileging statements from the Global North over the Global South, thus potentially exacerbating existing informational inequities. Regions such as Africa and the Middle East are at a disadvantage, with much lower factual accuracy. The performance fluctuations over time suggest that model updates may not consistently benefit all regions equally. Our study also offers insights into the impact of various LLM configuration settings, such as binary decision forcing, model re-runs and temperature, on model's factuality. Models constrained to binary (true/false) choices exhibit reduced factuality compared to those allowing an 'unclear' option. Single inference at a low temperature setting matches the reliability of majority voting across various configurations. The insights gained highlight the need for culturally diverse and geographically inclusive model training and evaluation. This approach is key to achieving global equity in technology, distributing AI benefits fairly worldwide.

Pretraining auto-regressive large language models (LLMs) with retrieval demonstrates better perplexity and factual accuracy by leveraging external databases. However, the size of existing pretrained retrieval-augmented LLM is still limited (e.g., Retro has 7.5B parameters), which limits the effectiveness of instruction tuning and zero-shot generalization. In this work, we introduce Retro 48B, the largest LLM pretrained with retrieval. Specifically, we continue to pretrain a 43B GPT model on additional 100 billion tokens using the Retro augmentation method by retrieving from 1.2 trillion tokens. Notably, the obtained foundation model, Retro 48B, largely outperforms the counterpart GPT 43B trained on 1.2T tokens in terms of perplexity with only 2.58% additional GPU hours, demonstrating the significant scaling potential of the method. After instruction tuning on Retro, InstructRetro demonstrates significant improvement over the instruction tuned GPT on a wide range of zero-shot tasks. Specifically, the average improvement of InstructRetro is 7% over its GPT counterpart across 8 short-form QA and reading comprehension tasks, 10% over GPT across 4 challenging long-form QA tasks, and 16% over GPT across 3 summarization tasks. Surprisingly, we find that one can ablate the encoder from InstructRetro architecture and directly use its decoder backbone, while achieving comparable results. Our results highlight the promising direction to obtain a better GPT decoder through continued pretraining with retrieval before instruction tuning. Our code and checkpoints are publicly available at: https://github.com/NVIDIA/Megatron-LM/tree/InstructRetro/tools/retro.

Hulu ran an anti-Hamas ad that appears to be made using artificial intelligence to show an idealized version of Gaza--claiming this paradise destination could exist if not for Hamas. The 30-second spot, opening like a tourism ad, shows palm trees and coastlines. There are five-star hotels and children playing. People dance, eat, and laugh, while a voiceover encourages visitors to "experience a culture rich in tradition." But it suddenly shifts, turning the face of a smiling man into a grimacing one.

The first reviews of Apple's Vision Pro headset, from publications with early access to the company's attempt to create the next computing platform, talk of a big leap forward for face-mounted computers, for better or worse. The US-only headset, first announced in June last year, aims to move "spatial computing" beyond the limited mixed-reality offered by rivals from Meta, Microsoft and others. It is packed with cutting-edge technology including 3D cameras on the front to capture videos, the ability to blend the real and virtual worlds with hand and eye tracking, plus a display on the front that shows a simulacrum of the wearer's eyes. But at a cost of 3,499 (about 2,760) in the US it has a lot of work to do to convince consumers and developers alike that it can be anything other than a super-expensive niche toy for tech enthusiasts. The Verge's Nilay Patel called the Vision Pro an "astounding product" but one with a lot of big trade-offs, including messing up your hair each time you put it on: "Apple is very proud of the displays inside the Vision Pro, and for good reason – they represent a huge leap forward in display technology," he wrote.