The Poisson equation is commonly encountered in engineering, including in computational fluid dynamics where it is needed to compute corrections to the pressure field. We propose a novel fully convolutional neural network (CNN) architecture to infer the solution of the Poisson equation on a 2D Cartesian grid of varying size and spacing given the right hand side term, arbitrary Dirichlet boundary conditions and grid parameters which provides unprecendented versatility in this application. The boundary conditions are handled using a novel approach by decomposing the original Poisson problem into a homogeneous Poisson problem plus four inhomogeneous Laplace sub-problems. The model is trained using a novel loss function approximating the continuous $L^p$ norm between the prediction and the target. Analytical test cases indicate that our CNN architecture is capable of predicting the correct solution of a Poisson problem with mean percentage errors of 15% and promises improvements in wall-clock runtimes for large problems. Furthermore, even when predicting on meshes denser than previously encountered, our model demonstrates encouraging capacity to reproduce the correct solution profile.

Artificial neural networks have exceeded human-level performance in accomplishing several individual tasks (e.g. voice recognition, object recognition, and video games). However, such success remains modest compared to human intelligence that can learn and perform an unlimited number of tasks. Humans' ability of learning and accumulating knowledge over their lifetime is an essential aspect of their intelligence. Continual machine learning aims at a higher level of machine intelligence through providing the artificial agents with the ability to learn online from a non-stationary and never-ending stream of data. A key component of such a never-ending learning process is to overcome the catastrophic forgetting of previously seen data, a problem that neural networks are well known to suffer from. The work described in this thesis has been dedicated to the investigation of continual learning and solutions to mitigate the forgetting phenomena in neural networks. To approach the continual learning problem, we first assume a task incremental setting where tasks are received one at a time and data from previous tasks are not stored. Since the task incremental setting can't be assumed in all continual learning scenarios, we also study the more general online continual setting. We consider an infinite stream of data drawn from a non-stationary distribution with a supervisory or self-supervisory training signal. The proposed methods in this thesis have tackled important aspects of continual learning. They were evaluated on different benchmarks and over various learning sequences. Advances in the state of the art of continual learning have been shown and challenges for bringing continual learning into application were critically identified.

In wondering what can be done to steer civilization away from the abyss, I confess to being increasingly puzzled by the central enigma of contemporary cognitive psychology: To what degree are we consciously capable of changing our minds? I don't mean changing our minds as to who is the best NFL quarterback, but changing our convictions about major personal and social issues that should unite but invariably divide us. As a senior neurologist whose career began before CAT and MRI scans, I have come to feel that conscious reasoning, the commonly believed remedy for our social ills, is an illusion, an epiphenomenon supported by age-old mythology rather than convincing scientific evidence. If so, it's time for us to consider alternate ways of thinking about thinking that are more consistent with what little we do understand about brain function. I'm no apologist for artificial intelligence, but if we are going to solve the world's greatest problems, there are several major advantages in abandoning the notion of conscious reason in favor of seeing humans as having an AI-like "black-box" intelligence. To believe that we can accurately determine whether or not consciousness contains causal properties is sheer folly. But first, a brief overview as to why I feel so strongly that purely conscious thought isn't physiologically likely.

The key question on the mind of policymakers now is whether Artificial Intelligence would be able to deliver on its promises instead of entering another season of scepticism and stagnation. The quest for Artificial Intelligence (AI) has travelled through multiple "seasons of hope and despair" since the 1950s. The introduction of neural networks and deep learning in late 1990s has generated a new wave of interest in AI and growing optimism in the possibility of applying it to a wide range of activities, including diplomacy. The key question on the mind of policymakers now is whether AI would be able to deliver on its promises instead of entering another season of scepticism and stagnation. This paper evaluates the potential of IA to provide reliable assistance in areas of diplomatic interest such as in consular services, crisis management, public diplomacy and international negotiations, as well as the ratio between costs and contributions of AI applications to diplomatic work.

Quantizing large Neural Networks (NN) while maintaining the performance is highly desirable for resource-limited devices due to reduced memory and time complexity. NN quantization is usually formulated as a constrained optimization problem and optimized via a modified version of gradient descent. In this work, by interpreting the continuous parameters (unconstrained) as the dual of the quantized ones, we introduce a Mirror Descent (MD) framework (Bubeck (2015)) for NN quantization. Specifically, we provide conditions on the projections (i.e., mapping from continuous to quantized ones) which would enable us to derive valid mirror maps and in turn the respective MD updates. Furthermore, we discuss a numerically stable implementation of MD by storing an additional set of auxiliary dual variables (continuous). This update is strikingly analogous to the popular Straight Through Estimator (STE) based method which is typically viewed as a "trick" to avoid vanishing gradients issue but here we show that it is an implementation method for MD for certain projections. Our experiments on standard classification datasets (CIFAR-10/100, TinyImageNet) with convolutional and residual architectures show that our MD variants obtain fully-quantized networks with accuracies very close to the floating-point networks.

Observations of traffic participants and their environment enable humans to drive road vehicles safely. However, when being driven, there is a notable difference between having a non-experienced vs. an experienced driver. One may get the feeling, that the latter one anticipates what may happen in the next few moments and considers these foresights in his driving behavior. To make the driving style of automated vehicles comparable to a human driver in the sense of comfort and perceived safety, the aforementioned anticipation skills need to become a built-in feature of self-driving vehicles. This article provides a systematic comparison of methods and strategies to generate this intention for self-driving cars using machine learning techniques. To implement and test these algorithms we use a large data set collected over more than 30000 km of highway driving and containing approximately 40000 real world driving situations. Moreover, we show that it is possible to certainly detect more than 47 % of all lane changes on German highways 3 or more seconds in advance with a false positive rate of less than 1 %. This enables us to predict the lateral position with a prediction horizon of 5 s with a median error of less than 0.21 m.

Generating complex multi-turn goal-oriented dialogue agents is a difficult problem that has seen a considerable focus from many leaders in the tech industry, including IBM, Google, Amazon, and Microsoft. This is in large part due to the rapidly growing market demand for dialogue agents capable of goal-oriented behaviour. Due to the business process nature of these conversations, end-to-end machine learning systems are generally not a viable option, as the generated dialogue agents must be deployable and verifiable on behalf of the businesses authoring them. In this work, we propose a paradigm shift in the creation of goal-oriented complex dialogue systems that dramatically eliminates the need for a designer to manually specify a dialogue tree, which nearly all current systems have to resort to when the interaction pattern falls outside standard patterns such as slot filling. We propose a declarative representation of the dialogue agent to be processed by state-of-the-art planning technology. Our proposed approach covers all aspects of the process; from model solicitation to the execution of the generated plans/dialogue agents. Along the way, we introduce novel planning encodings for declarative dialogue synthesis, a variety of interfaces for working with the specification as a dialogue architect, and a robust executor for generalized contingent plans. We have created prototype implementations of all components, and in this paper, we further demonstrate the resulting system empirically.

In late August 2019, researchers at Google released a paper titled Weight Agnostic Neural Networks, opening our eyes to a missing piece of the puzzle in our quest to create Artificial Intelligence (AI) as close as possible to natural brains: instinct. This article covers the ways of finding Artificial Neural Network (ANN) architectures until now, points out the importance of instinct in a brain and, for the detail-hungry, describes how instinct can be incorporated into AI. The architecture of an ANN, in a nutshell, refers to the number, arrangement and connections of neurons therein. When building an ANN to solve a problem, the best architecture possible under all constraints is desired. After all, it does not pay to have a sub-optimal solution.

The Weak Completion Semantics (WCS) is a computational cognitive theory that has shown to be successful in modeling episodes of human reasoning. As the WCS is a recently developed logic programming approach, this paper investigates the correspondence of the WCS with respect to the well-established Answer Set Semantics (ASP). The underlying three-valued logic of both semantics is different and their models are evaluated with respect to different program transformations. We first illustrate these differences by the formal representation of some examples of a well-known psychological experiment, the suppression task. After that, we will provide a translation from logic programs understood under the WCS into logic programs understood under the ASP. In particular, we will show that logic programs under the WCS can be represented as logic programs under the ASP by means of a definition completion, where all defined atoms in a program must be false when their definitions are false.

The first part of the literature review depicts the role AI technology has today as long as some of its risks and negative impacts on society. The table below (Figure 1.4) summarises the themes addressed. This review of the adverse consequences of AI on society only covers the implementation of automated decision systems. It does not encompass any of the issues related to the impact of social media platforms such as filter bubbles, echo chambers of public opinion, data privacy, mass surveillance or discriminatory ads. It also does not provide either any insights regarding the threat of misuse by bad actors or criminals and the danger of a jobless future.