OpenAI, the research laboratory behind AI program ChatGPT, has released a tool designed to detect whether text has been written by artificial intelligence, but warns it's not completely reliable – yet. In a blog post on Tuesday, OpenAI linked to a new classifier tool that has been trained to distinguish between text written by a human and that written by a variety of AI, not just ChatGPT. Open AI researchers said that while it was "impossible to reliably detect all AI-written text", good classifiers could pick up signs that text was written by AI. The tool could be useful in cases where AI was used for "academic dishonesty" and when AI chatbots were positioned as humans, they said. But they admited the classifier "is not fully reliable" and only correctly identified 26% of AI-written English texts.

Check out all the on-demand sessions from the Intelligent Security Summit here. Ray, the popular open-source machine learning (ML) framework, has released its 2.2 version with improved performance and observability capabilities, as well as features that can help to enable reproducibility. The Ray technology is widely used by organizations to scale ML models across clusters of hardware, for both training and inference. Among Ray's many users is generative AI pioneer OpenAI, which uses Ray to scale and enable a variety of workloads, including supporting ChatGPT. The lead commercial sponsor behind the Ray open-source technology is San Francisco-based Anyscale, which has raised $259 million in funding to date.

OpenAI said Tuesday that it has released an AI "classifier" for identifying AI-authored text written by AI like its own ChatGPT. The problem? ChatGPT is pretty good at evading OpenAI's new tool. ChatGPT has absolutely overwhelmed academia, where students are using it as a virtual assistant of sorts in a variety of tasks. Unfortunately, some students are crossing the line and using it to create content that they are passing off as original--cheating, in other words. The trouble is trying to determine which answers were written by a human, and which by an AI.

Currently, large pre-trained language models are widely applied in neural code completion systems. Though large code models significantly outperform their smaller counterparts, around 70% displayed code completions from Copilot are not accepted by developers. Being reviewed but not accepted, their help to developer productivity is considerably limited. Even worse, considering the high cost of the large code models, it is a huge waste of computing resources and energy. To fill this significant gap, we first investigate the prompts of unhelpful code completions, and empirically find four observable patterns that cause such prompts, all of which are inherent, namely, they can hardly be addressed by improving the accuracy of the model. This demonstrates the feasibility of identifying such prompts based on the prompts themselves. Motivated by this finding, we propose an early-rejection mechanism to turn down low-return prompts by foretelling the code completion qualities without sending them to the code completion system. Furthermore, we propose a lightweight Transformer-based estimator to demonstrate the feasibility of the mechanism. The experimental results show that the proposed estimator helps save 23.3% of computational cost measured in floating-point operations for the code completion systems, and 80.2% of rejected prompts lead to unhelpful completion

We examine climate-related disclosures in a large sample of reports published by banks that officially endorsed the recommendations of the Task Force for Climate-related Financial Disclosures (TCFD). In doing so, we introduce a new application of the zero-shot text classification. By developing a set of fine-grained TCFD labels, we show that zero-shot analysis is a useful tool for classifying climate-related disclosures without further model training. Overall, our findings indicate that corporate climate-related disclosures grew dynamically after the launch of the TCFD recommendations. However, there are marked differences in the extent of reporting by recommended disclosure topic, suggesting that some recommendations have not yet been fully met. Our findings yield important conclusions for the design of climate-related disclosure frameworks.

Recent years have witnessed a big convergence of language, vision, and multi-modal pretraining. In this work, we present mPLUG-2, a new unified paradigm with modularized design for multi-modal pretraining, which can benefit from modality collaboration while addressing the problem of modality entanglement. In contrast to predominant paradigms of solely relying on sequence-to-sequence generation or encoder-based instance discrimination, mPLUG-2 introduces a multi-module composition network by sharing common universal modules for modality collaboration and disentangling different modality modules to deal with modality entanglement. It is flexible to select different modules for different understanding and generation tasks across all modalities including text, image, and video. Empirical study shows that mPLUG-2 achieves state-of-the-art or competitive results on a broad range of over 30 downstream tasks, spanning multi-modal tasks of image-text and video-text understanding and generation, and uni-modal tasks of text-only, image-only, and video-only understanding. Notably, mPLUG-2 shows new state-of-the-art results of 48.0 top-1 accuracy and 80.3 CIDEr on the challenging MSRVTT video QA and video caption tasks with a far smaller model size and data scale. It also demonstrates strong zero-shot transferability on vision-language and video-language tasks. Code and models will be released in https://github.com/alibaba/AliceMind.

Follow this and additional works at: https://digitalcommons.unl.edu/libphilprac Subaveerapandiyan A Ph.D. Research Scholar Department of Library and Information Science Yenepoya (Deemed to be University), Mangalore, Karnataka, India Email: subaveerapandiyan@gmail.com ORCiD: https://orcid.org/0000-0002-2149-9897 Abstract This study aims to understand the perceptions and opinions of academicians towards ChatGPT-3 by collecting and analyzing social media comments, and a survey was conducted with library and information science professionals. The research uses a content analysis method and finds that while ChatGPT-3 can be a valuable tool for research and writing, it is not 100% accurate and should be cross-checked. The study also finds that while some academicians may not accept ChatGPT-3, most are starting to accept it. The study is beneficial for academicians, content developers, and librarians. Keywords: Conversational Generative Pre-training Transformer (ChatGPT), Artificial Intelligence in Academia, Academic Writing with ChatGPT, Library Services Introduction The OpenAI-developed GPT (Generative Pre-trained Transformer) model has a variation called ChatGPT. The GPT model was initially released in 2018 and trained using the Common Crawl, a sizable dataset of text from the internet. The Transformer design, revealed in a 2017 study by Google researchers, served as the model's foundation. Unsupervised learning was used to train the initial GPT model, which meant that it was trained on a sizable text dataset without any explicit labels or annotations.

Large language models can perform various reasoning tasks by using chain-of-thought prompting, which guides them to find answers through step-by-step demonstrations. However, the quality of the prompts depends on the demonstrations given to the models, and creating many of them by hand is costly. We introduce Synthetic prompting, a method that leverages a few handcrafted examples to prompt the model to generate more examples by itself, and selects effective demonstrations to elicit better reasoning. Our method alternates between a backward and forward process to generate new examples. The backward process generates a question that match a sampled reasoning chain, so that the question is solvable and clear. The forward process produces a more detailed reasoning chain for the question, improving the quality of the example. We evaluate our method on numerical, symbolic, and algorithmic reasoning tasks, and show that it outperforms existing prompting techniques.

By understanding the hidden locational clues in images, entirely new approaches of analyzing the natural and built environment are being opened up with profound implications for a number of fields, ranging from the recognition of weather, season, and climate patterns to rural and urban scene understanding, and improvements in navigation and self-driving car technology. Since the beginning of 2022, image geolocalization has additionally garnered extensive media coverage for becoming an immediate priority of investigative journalists and open source intelligence (OSINT) researchers in their attempt to verify information and to document war atrocities in Ukraine, extracting geolocational information from social media content. Despite high academic and public interest, image geolocalization remains an extremely challenging problem. This is because training datasets are geographically sparse, often limited to specific countries, and biased towards urban or rural scenes. The task is further complicated by the fact that geolocalization requires reasoning on multiple levels of geographic granularity (e.g.

Considerable advancements have been made in various NLP tasks based on the impressive power of large language models (LLMs) and many NLP applications are deployed in our daily lives. In this work, we challenge the capability of LLMs with the new task of Ethical Quandary Generative Question Answering. Ethical quandary questions are more challenging to address because multiple conflicting answers may exist to a single quandary. We explore the current capability of LLMs in providing an answer with a deliberative exchange of different perspectives to an ethical quandary, in the approach of Socratic philosophy, instead of providing a closed answer like an oracle. We propose a model that searches for different ethical principles applicable to the ethical quandary and generates an answer conditioned on the chosen principles through prompt-based few-shot learning. We also discuss the remaining challenges and ethical issues involved in this task and suggest the direction toward developing responsible NLP systems by incorporating human values explicitly.