Look at this: college students are sharing (anonymously) that they've started using AI tools to generate essays that can bypass anti-plagiarism software and score an A. The widespread use of the tools could reshape education and force schools to figure out new writing prompts or entirely fresh ways of assessing student performance to avoid being duped by the technology. Most projections have the AI niche reaching over $420 billion in total market size by 2028, a compound annual growth rate of 39.4 percent. Google is in talks to invest at least $200 million into AI start-up Cohere Inc., according to people familiar with the matter; another sign of the escalating arms race among large technology companies in the sector. There are some harbored fears surrounding AI ranging from doomsday scenarios to simple ethics concerns, but the overall trend is clear, and investors seem to have confidence that humanity will make the necessary adjustments to coexist with this new tech. Look at this: college students are sharing (anonymously) that they've started using AI tools to generate essays that can bypass anti-plagiarism software and score an A. The widespread use of the tools could reshape education and force schools to figure out new writing prompts or entirely fresh ways of assessing student performance to avoid being duped by the technology.

And coming to terms with "robot writing" might just improve writing instruction, S. Scott Graham writes. September 2022 was apparently the month artificial intelligence essay angst boiled over in academia, as various media outlets published opinion pieces lamenting the rise of AI writing systems that will ruin student writing and pave the way toward unprecedented levels of academic misconduct. Then, on Sept. 23, academic Twitter exploded into a bit of a panic on this topic. The firestorm was prompted by a post to the OpenAI subreddit where user Urdadgirl69 claimed to be getting straight A's with essays "written" using artificial intelligence. Professors on Reddit and Twitter alike expressed frustration and concern about how best to address the threat of AI essays.

"It doesn't feel like something I'd write, but it also doesn't not feel like something I'd write," a North Carolina State University student said about their work integrating prose from an artificial intelligence text-generating program into a final course essay. Paul Fyfe, associate professor of English and the student's instructor in the Data and the Human course, had asked students to "cheat" in this way and then reflect on how the experiment tested or changed their ideas about writing, AI or humanness. Humans have long relied on writing assistance powered by artificial intelligence to check spelling and grammar, predict text, translate or transcribe. Now, anyone with an internet connection can access an AI tool such as OpenAI or Moonbeam, give it a prompt and receive--in seconds--an essay written in humanlike prose. Instructors who are concerned that students will use these tools to cheat may hold fast to in-class writing assessments or install surveillance tools to try to detect misconduct. But others argue those are fools' errands.

Two Chinese entrepreneurs, Derek Li and Rick Chang, separately jumped into the country's tech boom nearly a decade ago, in the early days of Xi Jinping's rule. China's mobile-technology market was exploding, fueled by generous government subsidies and a light touch from regulators. Their respective businesses benefited greatly from the dynamism in China's tech sector, underpinned by Mr. Xi's push for innovation and entrepreneurship. That atmosphere has now soured, with the Chinese leader targeting what he calls the ills of unchecked capitalism. While he still lavishes support on strategic tech sectors, Mr. Xi has aimed a regulatory fusillade at "monopolistic" practices of internet giants and their handling of troves of citizen data.

The predictions of question answering (QA)systems are typically evaluated against manually annotated finite sets of one or more answers. This leads to a coverage limitation that results in underestimating the true performance of systems, and is typically addressed by extending over exact match (EM) with pre-defined rules or with the token-level F1 measure. In this paper, we present the first systematic conceptual and data-driven analysis to examine the shortcomings of token-level equivalence measures. To this end, we define the asymmetric notion of answer equivalence (AE), accepting answers that are equivalent to or improve over the reference, and publish over 23k human judgments for candidates produced by multiple QA systems on SQuAD. Through a careful analysis of this data, we reveal and quantify several concrete limitations of the F1 measure, such as a false impression of graduality, or missing dependence on the question. Since collecting AE annotations for each evaluated model is expensive, we learn a BERT matching (BEM) measure to approximate this task. Being a simpler task than QA, we find BEM to provide significantly better AE approximations than F1, and to more accurately reflect the performance of systems. Finally, we demonstrate the practical utility of AE and BEM on the concrete application of minimal accurate prediction sets, reducing the number of required answers by up to x2.6.

GitHub Copilot is an artificial intelligence model for automatically generating source code from natural language problem descriptions. Since June 2022, Copilot has officially been available for free to all students as a plug-in to development environments like Visual Studio Code. Prior work exploring OpenAI Codex, the underlying model that powers Copilot, has shown it performs well on typical CS1 problems thus raising concerns about the impact it will have on how introductory programming courses are taught. However, little is known about the types of problems for which Copilot does not perform well, or about the natural language interactions that a student might have with Copilot when resolving errors. We explore these questions by evaluating the performance of Copilot on a publicly available dataset of 166 programming problems. We find that it successfully solves around half of these problems on its very first attempt, and that it solves 60\% of the remaining problems using only natural language changes to the problem description. We argue that this type of prompt engineering, which we believe will become a standard interaction between human and Copilot when it initially fails, is a potentially useful learning activity that promotes computational thinking skills, and is likely to change the nature of code writing skill development.

Multimodal tasks in the fashion domain have significant potential for e-commerce, but involve challenging vision-and-language learning problems - e.g., retrieving a fashion item given a reference image plus text feedback from a user. Prior works on multimodal fashion tasks have either been limited by the data in individual benchmarks, or have leveraged generic vision-and-language pre-training but have not taken advantage of the characteristics of fashion data. Additionally, these works have mainly been restricted to multimodal understanding tasks. To address these gaps, we make two key contributions. First, we propose a novel fashion-specific pre-training framework based on weakly-supervised triplets constructed from fashion image-text pairs. We show the triplet-based tasks are an effective addition to standard multimodal pre-training tasks. Second, we propose a flexible decoder-based model architecture capable of both fashion retrieval and captioning tasks. Together, our model design and pre-training approach are competitive on a diverse set of fashion tasks, including cross-modal retrieval, image retrieval with text feedback, image captioning, relative image captioning, and multimodal categorization.

We introduce CS1QA, a dataset for code-based question answering in the programming education domain. CS1QA consists of 9,237 question-answer pairs gathered from chat logs in an introductory programming class using Python, and 17,698 unannotated chat data with code. Each question is accompanied with the student's code, and the portion of the code relevant to answering the question. We carefully design the annotation process to construct CS1QA, and analyze the collected dataset in detail. The tasks for CS1QA are to predict the question type, the relevant code snippet given the question and the code and retrieving an answer from the annotated corpus. Results for the experiments on several baseline models are reported and thoroughly analyzed. The tasks for CS1QA challenge models to understand both the code and natural language. This unique dataset can be used as a benchmark for source code comprehension and question answering in the educational setting.

Reconstructing force fields (FFs) from atomistic simulation data is a challenge since accurate data can be highly expensive. Here, machine learning (ML) models can help to be data economic as they can be successfully constrained using the underlying symmetry and conservation laws of physics. However, so far, every descriptor newly proposed for an ML model has required a cumbersome and mathematically tedious remodeling. We therefore propose using modern techniques from algorithmic differentiation within the ML modeling process -- effectively enabling the usage of novel descriptors or models fully automatically at an order of magnitude higher computational efficiency. This paradigmatic approach enables not only a versatile usage of novel representations and the efficient computation of larger systems -- all of high value to the FF community -- but also the simple inclusion of further physical knowledge such as higher-order information (e.g. Hessians, more complex partial differential equations constraints etc.), even beyond the presented FF domain.

Educational robots allow experimenting with a variety of principles from mechanics, electronics, and informatics. Here we propose ClipBot, a low-cost, do-it-yourself, robot whose skeleton is made of two paper clips. An Arduino nano microcontroller actuates two servo motors that move the paper clips. However, such mechanical configuration confers physical impairments to movement. This creates the need for and allows experimenting with artificial intelligence methods to overcome hardware limitations. We report our experience in the usage of this robot during the study week 'fascinating informatics', organized by the Swiss Foundation Schweizer Jugend Forscht (www.sjf.ch). Students at the high school level were asked to implement a genetic algorithm to optimize the movements of the robot until it learned to walk. Such a methodology allowed the robot to learn the motor actuation scheme yielding straight movement in the forward direction using less than 20 iterations.