The shelves are bare for consumer packaged goods (CPG) firms. Ongoing disruption to global supply chains continues to hurt food companies' ability to meet consumer demand for products. How might artificial intelligence (AI) helps CPG companies fight through the bottleneck? Most CPG companies are wrestling with a compelling challenge. The Coronavirus disease (COVID-19) pandemic is not going away anytime soon, if ever, which means that a CPG business must brace itself for ongoing disruptions caused by work slowdowns and outright shutdowns. For example, a factory in Malaysia was closed for 10 days because of a pandemic-related lockdown.

Got a moral quandary you don't know how to solve? Why not turn to the wisdom of artificial intelligence, aka Ask Delphi: an intriguing research project from the Allen Institute for AI that offers answers to ethical dilemmas while demonstrating in wonderfully clear terms why we shouldn't trust software with questions of morality. Ask Delphi was launched on October 14th, along with a research paper describing how it was made. From a user's point of view, though, the system is beguilingly simple to use. Just head to the website, outline pretty much any situation you can think of, and Delphi will come up with a moral judgement. Since Ask Delphi launched, its nuggets of wisdom have gone viral in news stories and on social media.

The design of widespread vision-and-language datasets and pre-trained encoders directly adopts, or draws inspiration from, the concepts and images of ImageNet. While one can hardly overestimate how much this benchmark contributed to progress in computer vision, it is mostly derived from lexical databases and image queries in English, resulting in source material with a North American or Western European bias. Therefore, we devise a new protocol to construct an ImageNet-style hierarchy representative of more languages and cultures. In particular, we let the selection of both concepts and images be entirely driven by native speakers, rather than scraping them automatically. Specifically, we focus on a typologically diverse set of languages, namely, Indonesian, Mandarin Chinese, Swahili, Tamil, and Turkish. On top of the concepts and images obtained through this new protocol, we create a multilingual dataset for {M}ulticultur{a}l {R}easoning over {V}ision and {L}anguage (MaRVL) by eliciting statements from native speaker annotators about pairs of images. The task consists of discriminating whether each grounded statement is true or false. We establish a series of baselines using state-of-the-art models and find that their cross-lingual transfer performance lags dramatically behind supervised performance in English. These results invite us to reassess the robustness and accuracy of current state-of-the-art models beyond a narrow domain, but also open up new exciting challenges for the development of truly multilingual and multicultural systems.

While large pre-trained models have enabled impressive results on a variety of downstream tasks, the largest existing models still make errors, and even accurate predictions may become outdated over time. Because detecting all such failures at training time is impossible, enabling both developers and end users of such models to correct inaccurate outputs while leaving the model otherwise intact is desirable. However, the distributed, black-box nature of the representations learned by large neural networks makes producing such targeted edits difficult. If presented with only a single problematic input and new desired output, fine-tuning approaches tend to overfit; other editing algorithms are either computationally infeasible or simply ineffective when applied to very large models. To enable easy post-hoc editing at scale, we propose Model Editor Networks with Gradient Decomposition (MEND), a collection of small auxiliary editing networks that use a single desired input-output pair to make fast, local edits to a pre-trained model. MEND learns to transform the gradient obtained by standard fine-tuning, using a low-rank decomposition of the gradient to make the parameterization of this transformation tractable. MEND can be trained on a single GPU in less than a day even for 10 billion parameter models; once trained MEND enables rapid application of new edits to the pre-trained model. Our experiments with T5, GPT, BERT, and BART models show that MEND is the only approach to model editing that produces effective edits for models with tens of millions to over 10 billion parameters. Increasingly large neural networks have become a fundamental tool in solving data-driven problems in computer vision (Huang et al., 2017) and natural language processing (Vaswani et al., 2017) in particular. However, a key challenge in deploying and maintaining such models is issuing patches to adjust model behavior after deployment (Sinitsin et al., 2020).

We study adaptive sensing of Cox point processes, a widely used model from spatial statistics. We introduce three tasks: maximization of captured events, search for the maximum of the intensity function and learning level sets of the intensity function. We model the intensity function as a sample from a truncated Gaussian process, represented in a specially constructed positive basis. In this basis, the positivity constraint on the intensity function has a simple form. We show how an minimal description positive basis can be adapted to the covariance kernel, non-stationarity and make connections to common positive bases from prior works. Our adaptive sensing algorithms use Langevin dynamics and are based on posterior sampling (\textsc{Cox-Thompson}) and top-two posterior sampling (\textsc{Top2}) principles. With latter, the difference between samples serves as a surrogate to the uncertainty. We demonstrate the approach using examples from environmental monitoring and crime rate modeling, and compare it to the classical Bayesian experimental design approach.

The robotic artist known as Ai-Da was scheduled to display her artwork alongside the great pyramids of Egypt on Thursday, though the show was nearly called off after both the robot and her human sculptor, Aidan Meller, were detained by Egyptian authorities for a week and a half until they could confirm that the artist was actually a spy. The incident began when border guards objected over Ai-da's camera eyes, which it uses in its creative process, and its on-board modem. "I can ditch the modems, but I can't really gouge her eyes out," Meller told The Guardian. The robot artist, which was built in 2019, typically travels via specialized cargo case and was held at the border until clearing customs on Wednesday evening, hours before the exhibit was scheduled to begin. "The British ambassador has been working through the night to get Ai-Da released, but we're right up to the wire now," Meller said, just before Ai-Da was sprung from robo-jail.

Following the successful use of online apps, tracking devices and telemedicine to control a pandemic, while enabling remote care, doctors in the Middle East have started using medtech gadgets for non-invasive as well as accurate procedures. Robots have been pitched as cops, office assistants and guides for the future, but the need for contactless healthcare has led to their adoption as support staff in hospitals, to monitor patients in quarantine. Robotic pharmacies have also become a feature at AI-backed facilities in the UAE, while droids have even conducted operations, one of which involved removal of a tumour. Considering the precision that smart machines bring to the table, robo-arms are now being used for removal and transfer of kidneys from donors to patients, at a hospital in Dubai. The surgeries that benefited two people were carried out using the state-of-the-art Da Vinci tool, which is a pair of robotic arms, that can imitate the movement of human hands.

Lily Xu is a PhD student at Harvard University, applying machine learning and game theory to wildlife conservation. She is particularly focused on the prevention of illegal wildlife poaching, and she told us about this interesting, and critically important, area of research. Green security is the challenge of environmental conservation under some unknown threat. The three domains that we focus on are illegal wildlife poaching, illegal logging and illegal fishing. Across all of these settings we have an environmental challenge, which is to preserve our natural ecosystems.

The Old Testament's'Tower of Babel' story is an origin myth that tries to explain why humanity doesn't speak a single, universal language. According to the Bible, a united human race that speaks the same language arrived in the land of Shinar and decided to build a tower tall enough to reach heaven. Annoyed -- once again, it can probably be said -- by humanity's growing arrogance and budding hubris, God confounded humanity's speech, dividing its people into separate linguistic groups that couldn't understand one another. Just to ensure they don't start comparing and contrasting their languages to reach some form of translating breakthrough, God dispersed humankind to all corners of the earth and set the stage for what is today a world of 6,500 languages. For God, a job well done and the situation remained static for centuries, that was until tribes starting trading with each other, armies started fighting one another, and diplomats initiated conflict resolution measures to try to end the wars that were often started due to misunderstandings of one kind or another.

After years of complaints that there were no rules to determine what constitutes a "sustainable" investment, investors are now fretting that there will soon be too many to navigate easily. More than 30 taxonomies outlining what is and isn't a green investment are being compiled by governments across Asia, Europe and Latin America, each one reflecting national economic idiosyncrasies that can jar with a global capital market that has seen trillions pour into sustainable funds. The European Union will introduce its green investment taxonomy, or common framework, in January to help asset managers inside the bloc and make green activities more visible and attractive to investors. The rules also aim to stamp out "green washing," whereby organizations overstate their environmental credentials. The U.K., which hosts the COP26 climate change conference from Oct. 31, is set to finalize its own taxonomy next year but has already signaled it will not just replicate what is drawn up across the channel.