I have been thinking of interactive ways of getting my postgraduate thesis on Racial Bias, Gender Bias, AI new ways to approach Human Computer Interaction out to everyone. Life has been super busy so I have decided to add snippets of the thesis for now. For this research paper, the researcher has identified a number of areas of concern in regards to systems powered by AI being deployed in situations that affect the lives of humans. These examples will be used to further highlight this area of concern. Suggestions have made that decision-support systems powered by AI can be used to augment human judgement and reduce both conscious and unconscious biases (Anderson & Anderson, 2007).

The Malta Information Technology Law Association (MITLA) has appealed to the government not to cede control of information to Artificial Intelligence (AI) and to ensure that information remains driven by humans. In a reaction document to Government's call for public consultation on an ethical framework for AI, MITLA recommended that the human-centric element be kept strongly in focus. To this end, MITLA proposed that human-operated'kill-switches' are introduced to over-ride AI "…malfunctioning or worse, of AI systems". MITLA made a strong recommendation to Government not to grant distinct legal personality to AI systems, so as not to undermine the fundamental principle of human-centric AI, which should respect fundamental human rights. MITLA also said that to ensure that the process of creating an ethical AI framework is coherent, digital rights should first be introduced into the Maltese Constitution.

The Joint Artificial Intelligence Center, stood up simply a year ago, has plans to contract an ethicist to help manage the Defense Department's advancement and use of artificial intelligence advances. "One of the positions we are going to fill will be someone who isn't simply taking a gander at specialized principles, yet who is an ethicist," said Air Force Lt. Gen. Jack Shanahan, the JAIC's chief. "We will get somebody who will have a profound foundation in morals, and afterward the attorneys inside the office will take a gander at how we really heat this into the Department of Defense." Speaking Aug. 30 at the Pentagon, Shanahan gave a report on the JAIC, where he's been since January. He'd recently driven Project Maven, an artificial intelligence AI pathfinder venture under the undersecretary of protection for intelligence.

Decades of science fiction have primed people to conjure a very specific image when the subject of artificial intelligence arises: The malevolent robotic overlord that has evolved beyond the need for humans, or concern for their safety. And with that image in mind, it's easy to imagine that it's still a long way off. But in actuality, AI is already here, and its most common form is a lot closer to home than many people realize. "It's already hard to interact with a modern technology that doesn't in some way benefit from AI, whether it's obvious or not," said Mason McDaniel, chief technology officer at the Bureau of Alcohol, Tobacco, Firearms and Explosives, said during a Sept. 4 Meritalk webinar. "And I'd say that's going to continue as we move forward. And AI is going to increasingly be embedded and hidden behind the scenes simply as part of the applications; you're not going to go specifically to an AI app." "And I see that as those mature, as those get better, those are going to become more of a primary interface, whether it's spoken or using natural language text typing, to ask the devices for what you want," McDaniel said.

In remote sensing, it is often difficult to acquire or collect a large dataset that is accurately labeled. This difficulty is often due to several issues including but not limited to the study site's spatial area and accessibility, errors in global positioning system (GPS), and mixed pixels caused by an image's spatial resolution. An approach, with two variations, is proposed that estimates multiple target signatures from mixed training samples with imprecise labels: Multi-Target Multiple Instance Adaptive Cosine Estimator (Multi-Target MI-ACE) and Multi-Target Multiple Instance Spectral Match Filter (Multi-Target MI-SMF). The proposed methods address the problems above by directly considering the multiple-instance, imprecisely labeled dataset and learns a dictionary of target signatures that optimizes detection using the Adaptive Cosine Estimator (ACE) and Spectral Match Filter (SMF) against a background. The algorithms have two primary steps, initialization and optimization. The initialization process determines diverse target representatives, while the optimization process simultaneously updates the target representatives to maximize detection while learning the number of optimal signatures to describe the target class. Three designed experiments were done to test the proposed algorithms: a simulated hyperspectral dataset, the MUUFL Gulfport hyperspectral dataset collected over the University of Southern Mississippi-Gulfpark Campus, and the AVIRIS hyperspectral dataset collected over Santa Barbara County, California. Both simulated and real hyperspectral target detection experiments show the proposed algorithms are effective at learning target signatures and performing target detection.

In this work we propose a new method to optimize the architecture of an artificial neural network. The algorithm proposed, called Greedy Search for Neural Network Architecture, aims to minimize the complexity of the architecture search and the complexity of the final model selected without compromising the predictive performance. The reduction of the computational cost makes this approach appealing for two reasons. Firstly, there is a need from domain scientists to easily interpret predictions returned by a deep learning model and this tends to be cumbersome when neural networks have complex structures. Secondly, the use of neural networks is challenging in situations with compute/memory limitations. Promising numerical results show that our method is competitive against other hyperparameter optimization algorithms for attainable performance and computational cost. We also generalize the definition of adjusted score from linear regression models to neural networks. Numerical experiments are presented to show that the adjusted score can boost the greedy search to favor smaller architectures over larger ones without compromising the predictive performance.

We contribute the largest publicly available dataset of naturally occurring factual claims for the purpose of automatic claim verification. It is collected from 26 fact checking websites in English, paired with textual sources and rich metadata, and labelled for veracity by human expert journalists. We present an in-depth analysis of the dataset, highlighting characteristics and challenges. Further, we present results for automatic veracity prediction, both with established baselines and with a novel method for joint ranking of evidence pages and predicting veracity that outperforms all baselines. Significant performance increases are achieved by encoding evidence, and by modelling metadata. Our best-performing model achieves a Macro F1 of 49.2%, showing that this is a challenging testbed for claim veracity prediction.

Most research on lifelong learning (LLL) applies to images or games, but not language. Here, we introduce LAMAL, a simple yet effective method for LLL based on language modeling. LAMAL replays pseudo samples of previous tasks while requiring no extra memory or model capacity. To be specific, LAMAL is a language model learning to solve the task and generate training samples at the same time. At the beginning of training a new task, the model generates some pseudo samples of previous tasks to train alongside the data of the new task. The results show that LAMAL prevents catastrophic forgetting without any sign of intransigence and can solve up to five very different language tasks sequentially with only one model. Overall, LAMAL outperforms previous methods by a considerable margin and is only 2-3\% worse than multitasking which is usually considered as the upper bound of LLL. Our source code is available at https://github.com/xxx.

Artificial intelligence (AI) issues in intellectual property are becoming increasingly ubiquitous. For example, the US Patent and Trademark Office (PTO) has issued "thousands of patents on AI technologies."1 The Persado Message Machine, which creates written content by way of data science and AI and is used by "[o]ver 250 of the world's most valuable brands," can generate marketing messages in 25 languages.2 And the University of Surrey in the UK just filed two patent applications--one that claims a "beverage container based on fractal geometry" and one that claims a device "that may help with search and rescue operations"--alleged to be the first inventions "created autonomously by artificial intelligence (AI) without a human inventor."3 Perhaps given these developments, the PTO has decided that the time is now to begin asking questions that broadly address how AI shifts our basic understanding of patent law concepts like inventorship, eligibility, enablement, and the level of ordinary skill in the art.

Named one of the "Top Ten Technology Books of 2018" by Peter High, Forbes.com "Compared with the amount of ink spilled over the prospects of artificial general intelligence and all its accompanying fears--the singularity!--there's "…a readily understandable guide to artificial intelligence and the immensely consequential effects it could have on our economy, our society and our political system." -- Robert E. Rubin, former U.S. Treasury secretary and co-chair Emeritus, Council on Foreign Relations This 2018 book…on the timely topic of AI - tops my summer reading list. "This is a timely book, well written, and accessible putting forward their insights, and is well worth reading." Lawrence H. Summers, Charles W. Eliot Professor, former president, Harvard University; former secretary, US Treasury; and former chief economist, World Bank-- "AI may transform your life.