Controllable text generation concerns two fundamental tasks of wide applications, namely generating text of given attributes (i.e., attribute-conditional generation), and minimally editing existing text to possess desired attributes (i.e., text attribute transfer). Extensive prior work has largely studied the two problems separately, and developed different conditional models which, however, are prone to producing biased text (e.g., various gender stereotypes). This paper proposes to formulate controllable text generation from a principled causal perspective which models the two tasks with a unified framework. A direct advantage of the causal formulation is the use of rich causality tools to mitigate generation biases and improve control. We treat the two tasks as interventional and counterfactual causal inference based on a structural causal model, respectively. We then apply the framework to the challenging practical setting where confounding factors (that induce spurious correlations) are observable only on a small fraction of data. Experiments show significant superiority of the causal approach over previous conditional models for improved control accuracy and reduced bias.

Does the recent flurry of headlines about Facebook and the negative outcomes produced by its algorithms have you worried about the future and the implications of widespread AI usage? It's a rational response to have during an alarming news cycle. However, this situation shouldn't be interpreted as a death knell for the use of AI in human communications. It's more of a cautionary example of the disastrous consequences that can occur as a result of not using AI in a responsible way. Read on to learn more about ethical technology, data quality, and the significance of human-in-the-loop AI.

As artificial intelligence plays an increasingly important role in the R&D process, the premise that invention is a uniquely human characteristic is being challenged. Patent offices and courts around the world have recently been grappling with the question of whether an AI system can be the inventor of a patent. This has been prompted by Dr. Stephen Thaler's applications to designate his AI system (known as'DABUS') as the inventor of patents filed in multiple jurisdictions. Most recently, the appeal board of the European Patent Office (EPO) refused Dr. Thaler's patent applications because there was no valid inventor. Dr. Thaler, as part of the Artificial Inventor Project, is pursuing parallel patent applications across over fifteen jurisdictions which designate his AI system, DABUS, as the inventor.

Gibson Dunn lawyers examine a number of the most significant developments from 2021 affecting companies as they navigate the evolving AI and autonomous systems landscape.

Knowledge graphs (KGs) have become a valuable asset for many AI applications. Although some KGs contain plenty of facts, they are widely acknowledged as incomplete. To address this issue, many KG completion methods are proposed. Among them, open KG completion methods leverage the Web to find missing facts. However, noisy data collected from diverse sources may damage the completion accuracy. In this paper, we propose a new trustworthy method that exploits facts for a KG based on multi-sourced noisy data and existing facts in the KG. Specifically, we introduce a graph neural network with a holistic scoring function to judge the plausibility of facts with various value types. We design value alignment networks to resolve the heterogeneity between values and map them to entities even outside the KG. Furthermore, we present a truth inference model that incorporates data source qualities into the fact scoring function, and design a semi-supervised learning way to infer the truths from heterogeneous values. We conduct extensive experiments to compare our method with the state-of-the-arts. The results show that our method achieves superior accuracy not only in completing missing facts but also in discovering new facts.

Even after more than a hundred years after its introduction, histology remains the gold standard in tumor diagnosis and prognosis. Anatomic pathologists evaluate histology to stratify cancer patients into different groups depending on their tumor genotypes and phenotypes, and their clinical outcome [1,2]. However, human evaluation of histological slides is subjective and not repeatable [3]. Furthermore, histological assessment is a time-consuming process that requires highly trained professionals. With significant technological advances in the last decade, techniques such as whole slide imaging (WSI) and deep learning (DL) are now widely available.

From better healthcare access to improved food security, machine learning could tackle a wide range of challenges in developing countries. In 2020, a study published in Nature showed that Google's machine learning artificial intelligence programme, DeepMind AI, outperformed radiologists in detecting breast cancer. After being trained on thousands of mammograms, the system was able to accurately identify 89% of breast cancer cases, compared to radiologists' 74%. Just imagine what a difference the deployment of such a system could make in sub-Saharan Africa, where there are 0.2 doctors per 1000 people, according to the World Bank. Marilyn Moodley, Country Leader for South Africa and WECA (West, East, Central Africa) at SoftwareONE, says machine learning can help with some of the region's most pervasive problems, from reducing poverty and improving education to delivering healthcare and addressing sustainability challenges such as food demand.

With every advancement in the digital world, we unlock a limitless resource: data. It is both a by-product and a driver of global development that has transformed how we make decisions. Not only do we have increased granularity and accuracy to inform evidence-based decision making, but through AI and machine learning, we enable technology to make decisions on our behalf. The value of this data is well established in the private sector. Successful businesses have captured this value through increasingly efficient and targeted advertisements and product design – with the global marketing data market worth an estimated at $52 billion in 2021.

Value-function (VF) approximation is a central problem in Reinforcement Learning (RL). Classical non-parametric VF estimation suffers from the curse of dimensionality. As a result, parsimonious parametric models have been adopted to approximate VFs in high-dimensional spaces, with most efforts being focused on linear and neural-network-based approaches. Differently, this paper puts forth a a parsimonious non-parametric approach, where we use stochastic low-rank algorithms to estimate the VF matrix in an online and model-free fashion. Furthermore, as VFs tend to be multi-dimensional, we propose replacing the classical VF matrix representation with a tensor (multi-way array) representation and, then, use the PARAFAC decomposition to design an online model-free tensor low-rank algorithm. Different versions of the algorithms are proposed, their complexity is analyzed, and their performance is assessed numerically using standardized RL environments.

In this paper, we create EMIR, the first-ever Music Information Retrieval dataset for Ethiopian music. EMIR is freely available for research purposes and contains 600 sample recordings of Orthodox Tewahedo chants, traditional Azmari songs and contemporary Ethiopian secular music. Each sample is classified by five expert judges into one of four well-known Ethiopian Kinits, Tizita, Bati, Ambassel and Anchihoye. Each Kinit uses its own pentatonic scale and also has its own stylistic characteristics. Thus, Kinit classification needs to combine scale identification with genre recognition. After describing the dataset, we present the Ethio Kinits Model (EKM), based on VGG, for classifying the EMIR clips. In Experiment 1, we investigated whether Filterbank, Mel-spectrogram, Chroma, or Mel-frequency Cepstral coefficient (MFCC) features work best for Kinit classification using EKM. MFCC was found to be superior and was therefore adopted for Experiment 2, where the performance of EKM models using MFCC was compared using three different audio sample lengths. 3s length gave the best results. In Experiment 3, EKM and four existing models were compared on the EMIR dataset: AlexNet, ResNet50, VGG16 and LSTM. EKM was found to have the best accuracy (95.00%) as well as the fastest training time. We hope this work will encourage others to explore Ethiopian music and to experiment with other models for Kinit classification.