Text-based games present a unique class of sequential decision making problem in which agents interact with a partially observable, simulated environment via actions and observations conveyed through natural language. Such observations typically include instructions that, in a reinforcement learning (RL) setting, can directly or indirectly guide a player towards completing reward-worthy tasks. In this work, we study the ability of RL agents to follow such instructions. We conduct experiments that show that the performance of state-of-the-art text-based game agents is largely unaffected by the presence or absence of such instructions, and that these agents are typically unable to execute tasks to completion. To further study and address the task of instruction following, we equip RL agents with an internal structured representation of natural language instructions in the form of Linear Temporal Logic (LTL), a formal language that is increasingly used for temporally extended reward specification in RL. Our framework both supports and highlights the benefit of understanding the temporal semantics of instructions and in measuring progress towards achievement of such a temporally extended behaviour. Experiments with 500+ games in TextWorld demonstrate the superior performance of our approach.

Drones have shown to be useful aerial vehicles for unmanned transport missions such as food and medical supply delivery. This can be leveraged to deliver life-saving nutrition and medicine for people in emergency situations. However, commercial drones can generally only carry 10 % - 30 % of their own mass as payload, which limits the amount of food delivery in a single flight. One novel solution to noticeably increase the food-carrying ratio of a drone, is recreating some structures of a drone, such as the wings, with edible materials. We thus propose a drone, which is no longer only a food transporting aircraft, but itself is partially edible, increasing its food-carrying mass ratio to 50 %, owing to its edible wings. Furthermore, should the edible drone be left behind in the environment after performing its task in an emergency situation, it will be more biodegradable than its non-edible counterpart, leaving less waste in the environment. Here we describe the choice of materials and scalable design of edible wings, and validate the method in a flight-capable prototype that can provide 300 kcal and carry a payload of 80 g of water.

Recent advances in large-scale pre-training provide large models with the potential to learn knowledge from the raw text. It is thus natural to ask whether it is possible to leverage these large models as knowledge bases for downstream tasks. In this work, we answer the aforementioned question in unsupervised knowledge-grounded conversation. We explore various methods that best elicit knowledge from large models. Our human study indicates that, though hallucinations exist, large models post the unique advantage of being able to output common sense and summarize facts that cannot be directly retrieved from the search engine. To better exploit such generated knowledge in dialogue generation, we treat the generated knowledge as a noisy knowledge source and propose the posterior-based reweighing as well as the noisy training strategy. Empirical results on two benchmarks show advantages over the state-of-the-art methods.

Large pretrained language models generate fluent text but are notoriously hard to controllably sample from. In this work, we study constrained sampling from such language models: generating text that satisfies user-defined constraints, while maintaining fluency and the model's performance in a downstream task. We propose MuCoLa -- a sampling procedure that combines the log-likelihood of the language model with arbitrary (differentiable) constraints in a single energy function, and then generates samples in a non-autoregressive manner. Specifically, it initializes the entire output sequence with noise and follows a Markov chain defined by Langevin Dynamics using the gradients of the energy function. We evaluate MuCoLa on text generation with soft and hard constraints as well as their combinations obtaining significant improvements over competitive baselines for toxicity avoidance, sentiment control, and keyword-guided generation.

Word alignment is to find translationally equivalent words between source and target sentences. Previous work has demonstrated that self-training can achieve competitive word alignment results. In this paper, we propose to use word alignments generated by a third-party word aligner to supervise the neural word alignment training. Specifically, source word and target word of each word pair aligned by the third-party aligner are trained to be close neighbors to each other in the contextualized embedding space when fine-tuning a pre-trained cross-lingual language model. Experiments on the benchmarks of various language pairs show that our approach can surprisingly do self-correction over the third-party supervision by finding more accurate word alignments and deleting wrong word alignments, leading to better performance than various third-party word aligners, including the currently best one. When we integrate all supervisions from various third-party aligners, we achieve state-of-the-art word alignment performances, with averagely more than two points lower alignment error rates than the best third-party aligner. We released our code at https://github.com/sdongchuanqi/Third-Party-Supervised-Aligner.

Last week, MIT Technology Review brought together some of the world's sharpest minds dedicated to developing the technologies that are changing the way we live. EmTech, our annual flagship event covering cutting-edge developments and global trends, heard from experts working in fields as diverse as space commercialization to CRISPR gene editing, helping to set the agenda for the year ahead, and beyond. A massive thank you to everyone who attended in person and online! Kiran Musunuru, a top American cardiologist, is pioneering the use of gene editing to treat heart disease. He sat down with Antonio Regalado, our senior biotech writer, to discuss the clinical trial he's been overseeing to assess whether tweaking a cholesterol-regulating gene could help to prevent future deaths from heart disease.

The Cognitive Software leverages automation, big data scalability, speed, accuracy, and consistency for improved network optimization. The AI-based Cognitive Software solution also contributes to reducing carbon dioxide emissions from operational activities, for example, through the use of virtual drive-testing and remote automatic spectrum analysis. Additionally, stc Group has deployed 5G AI root-cause analysis capabilities to enable a better 5G experience for its subscribers. This future-proof deployment enables stc Group to leverage the Ericsson Performance Optimizers portfolio for surgical optimization analysis and recommendation. Ericsson Performance Optimizers use digital twin technology and advanced AI techniques like deep reinforcement learning and expert recommender systems to proactively provide mobile network optimization recommendations and resolve specific network performance issues, enabling a superior subscriber experience, while reducing operating costs.

This post contains a list of the AI-related seminars that are scheduled to take place between 7 November 2022 and 31 December 2022. All events detailed here are free and open for anyone to attend virtually. Does chocolate really cure cancer? Advances and Challenges in Conformal Prediction Speaker: Ryan Tibshirani Organised by: Harvard ML Theory Join the mailing list to find out how to access the seminars. Title to be confirmed Speaker: Tim G. J. Rudner (New York University) Organised by: New York University Please contact the organisers here if you are interested in attending the virtual seminar.

Google on Wednesday said it wanted to develop artificial intelligence using the world's one thousand most spoken languages as tech giants compete to dominate the internet's next battleground. Data is crucial to advances in AI, and Google and its big tech rivals want to tap information to help make products perform better and be more available to the widest possible audience. "Imagine a new internet user in Africa speaking Wolof... using their phone to ask where is the nearest pharmacy," said Johan Schalkwyk, a researcher at Google. Such situations "we take for granted," Schalkwyk told reporters, adding that languages were "not available to everyone in the world." According to Schalkwyk, there are more than 7,000 languages globally. However, Google only offers its translations for a little more than 130 of them.

OpenAI's DALL-E produced this image when prompted with the title of this article ("The Biggest ... [ ] Opportunity In Generative AI Is Language, Not Images"). The buzz around generative AI today is deafening. Generative AI refers to artificial intelligence that can generate novel content, rather than simply analyzing or acting on existing data. No topic in the world of technology is attracting more attention and hype right now. The white-hot epicenter of today's generative AI craze has been text-to-image AI. Text-to-image AI models generate detailed original images based on simple written inputs. The most well-known of these models include Stable Diffusion, Midjourney and OpenAI's DALL-E.