Large language models (LLMs) are capable of answering knowledge-intensive complex questions with chain-of-thought (CoT) reasoning. However, they tend to generate factually incorrect reasoning steps when the required knowledge is not available or up-to-date in models' parameters. Recent works turn to retrieving external knowledge to augment CoT reasoning. Despite being promising, these chain-based methods suffer from: 1) Negative retrieval. Unnecessary or incorrect retrieval may mislead the reasoning; 2) Limited sight. Lacking the ability to look backward or forward, a local error in one step will propagate along the chain. In this paper, we propose a novel approach: Probabilistic Tree-of-thought Reasoning (ProbTree). First, LLMs translate a complex question into a query tree, in which each non-root node denotes a sub-question of its parent node. Then, probabilistic reasoning is conducted over the tree, by solving questions from leaf to root considering the confidence of both question decomposing and answering. During reasoning, for leaf nodes, LLMs choose a more confident answer from Closed-book QA that employs parametric knowledge and Open-book QA that employs retrieved external knowledge, thus eliminating the negative retrieval problem. For non-leaf nodes, with the hierarchical structure, LLMs have broader sights and are able to globally reason with the information from child nodes, thus recovering from local errors. The experiments on three Complex QA datasets under the open-domain setting show that our approach outperforms SOTA methods significantly, demonstrating the effect of probabilistic tree-of-thought reasoning.

The turmoil at OpenAI over the past five days has captivated the tech industry and kept entrepreneurs, journalists, and anyone who still has an X account glued to their timelines for the latest emoji updates and lower-case missives. In the meantime, some of the most prominent AI companies--including OpenAI--continued to do what Silicon Valley is known for: Drop new products. The unexpected firing of Sam Altman, OpenAI's CEO, was followed by an avalanche of new AI features from competitors, including Anthropic and Stable Diffusion. On Tuesday afternoon, in the midst of turmoil, OpenAI rolled out ChatGPT with voice capabilities for free to all users. OpenAI had pre-released this in late September, but only for paid users.

In November 2022, OpenAI launched ChatGPT, a consumer-facing artificial intelligence tool that could hold a conversation with users, answer questions, and generate anything from poems to computer code to health advice. The initial technology was not perfect – it would sometimes "hallucinate", producing convincing but inaccurate information – but its potential generated enormous attention. A year later, ChatGPT's popularity has continued, with 100 million people using it on a weekly basis, and over 92% of Fortune 500 companies and several competitor firms looking to cash in or improve on the technology. But that's not why ChatGPT's creator, OpenAI, was in the news this week. Instead, OpenAI was the center of a fierce philosophical debate about what it means to develop artificial general intelligence for the benefit of humanity. To understand the current debate and its stakes requires going back to OpenAI's founding in December 2015.

Art created using generated Artificial Intelligence has taken the world by storm and generated excitement for many digital creators and technologists. However, the reception and reaction from artists have been mixed. Concerns about plagiarizing their artworks and styles for datasets and uncertainty around the future of digital art sparked movements in artist communities shunning the use of AI for generating art and protecting artists' rights. Collaborating with these tools for novel creative use cases also sparked hope from some creators. Artists are an integral stakeholder in the rapidly evolving digital creativity industry and understanding their concerns and hopes inform responsible development and use of creativity support tools. In this work, we study artists' sentiments about AI-generated art. We interviewed 7 artists and analyzed public posts from artists on social media platforms Reddit, Twitter and Artstation. We report artists' main concerns and hopes around AI-generated artwork, informing a way forward for inclusive development of these tools.

Language models produce a distribution over the next token; can we use this to recover the prompt tokens? We consider the problem of language model inversion and show that next-token probabilities contain a surprising amount of information about the preceding text. Often we can recover the text in cases where it is hidden from the user, motivating a method for recovering unknown prompts given only the model's current distribution output. We consider a variety of model access scenarios, and show how even without predictions for every token in the vocabulary we can recover the probability vector through search. On Llama-2 7b, our inversion method reconstructs prompts with a BLEU of 59 and token-level F1 of 78 and recovers 27% of prompts exactly. Language models are autoregressive, outputting the probability of each next token in a sequence conditioned on the preceeding text. This distribution is used to generate future tokens in the sequence. Can this distribution also be used to reconstruct the prompt? In most contexts, this question is pointless, since we have already conditioned on this information. However, increasingly language models are being offered "as a service" where the user may have access to the outputs, but not all of the true prompt. In this context, it may be of interest to users to know the prompt and, perhaps, for the service provider to protect it. This goal has been the focus of "jailbreaking" approaches that attempt to use the forward text generation of the model to reveal the prompt. We formalize this problem of prompt reconstruction as language model inversion, recovering the input prompt conditioned on the language model's next-token probabilities. Interestingly, work in computer vision has shown that probability predictions of image classifiers retain a surprising amount of detail (Dosovitskiy & Brox, 2016), so it is plausible that this also holds for language models. We propose an architecture that predicts prompts by"unrolling" the distribution vector into a sequence that can be processed effectively by a pretrained encoder-decoder language model. This method shows for the first time that language model predictions are mostly invertible: in many cases, we are able to recover very similar inputs to the original, sometimes getting the input text back exactly.

Video semantic segmentation is a pivotal aspect of video representation learning. However, significant domain shifts present a challenge in effectively learning invariant spatio-temporal features across the labeled source domain and unlabeled target domain for video semantic segmentation. To solve the challenge, we propose a novel DA-STC method for domain adaptive video semantic segmentation, which incorporates a bidirectional multi-level spatio-temporal fusion module and a category-aware spatio-temporal feature alignment module to facilitate consistent learning for domain-invariant features. Firstly, we perform bidirectional spatio-temporal fusion at the image sequence level and shallow feature level, leading to the construction of two fused intermediate video domains. This prompts the video semantic segmentation model to consistently learn spatio-temporal features of shared patch sequences which are influenced by domain-specific contexts, thereby mitigating the feature gap between the source and target domain. Secondly, we propose a category-aware feature alignment module to promote the consistency of spatio-temporal features, facilitating adaptation to the target domain. Specifically, we adaptively aggregate the domain-specific deep features of each category along spatio-temporal dimensions, which are further constrained to achieve cross-domain intra-class feature alignment and inter-class feature separation. Extensive experiments demonstrate the effectiveness of our method, which achieves state-of-the-art mIOUs on multiple challenging benchmarks. Furthermore, we extend the proposed DA-STC to the image domain, where it also exhibits superior performance for domain adaptive semantic segmentation. The source code and models will be made available at \url{https://github.com/ZHE-SAPI/DA-STC}.

Explainable AI (XAI) aims to provide insight into opaque model reasoning to humans and as such is an interdisciplinary field by nature. In this paper, we interviewed 10 practitioners to understand the possible usability of training data attribution (TDA) explanations and to explore the design space of such an approach. We confirmed that training data quality is often the most important factor for high model performance in practice and model developers mainly rely on their own experience to curate data. End-users expect explanations to enhance their interaction with the model and do not necessarily prioritise but are open to training data as a means of explanation. Within our participants, we found that TDA explanations are not well-known and therefore not used. We urge the community to focus on the utility of TDA techniques from the human-machine collaboration perspective and broaden the TDA evaluation to reflect common use cases in practice.

Altman's sudden move to join Microsoft is not finalized, Satya Nadella, CEO of Microsoft, signaled in an interview with CNBC on Monday. A person familiar with the matter said he would only return to OpenAI if the board members who ousted him stepped down. In the CNBC interview on Monday afternoon, Nadella sought to assure customers and investors that his company was on solid ground no matter the outcome. He left the door open for Altman to return to OpenAI or continue on as an AI leader at Microsoft, even though he announced late Sunday night that Altman was coming to Microsoft. "I'm open to both options," Nadella said in the interview with CNBC.

By the time I sat down with Christopher Nolan in his posh hotel suite not far from the White House, I guessed that he was tired of Washington, D.C. The day before, he'd toured the Oval Office and had lunch on Capitol Hill. Later that night, I'd watched him receive an award from the Federation for American Scientists, an organization that counts Robert Oppenheimer, the subject of Nolan's most recent film, among its founders. He'd endured a joke, repeated too many times by Senate Majority Leader Chuck Schumer, about the subject of his next film--"It's another biopic: Schumer." The award was sitting on an end table next to Nolan, who was dressed in brown slacks, a gray vest, and a navy suit jacket--his Anglo-formality undimmed by decades spent living in Los Angeles. "It's heavy, and glass, and good for self-defense," he said of the award, while filling his teacup.

We present GPQA, a challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. We ensure that the questions are high-quality and extremely difficult: experts who have or are pursuing PhDs in the corresponding domains reach 65% accuracy (74% when discounting clear mistakes the experts identified in retrospect), while highly skilled non-expert validators only reach 34% accuracy, despite spending on average over 30 minutes with unrestricted access to the web (i.e., the questions are "Google-proof"). The questions are also difficult for state-of-the-art AI systems, with our strongest GPT-4 based baseline achieving 39% accuracy. If we are to use future AI systems to help us answer very hard questions, for example, when developing new scientific knowledge, we need to develop scalable oversight methods that enable humans to supervise their outputs, which may be difficult even if the supervisors are themselves skilled and knowledgeable. The difficulty of GPQA both for skilled non-experts and frontier AI systems should enable realistic scalable oversight experiments, which we hope can help devise ways for human experts to reliably get truthful information from AI systems that surpass human capabilities.