BANKS have been rolling out chatbots, which allow users to bank without interacting with a human. Banking giant Citi is joining the mix, planning to launch its own chatbot in Hong Kong, except it's on (Facebook) Messenger. This makes sense for Citi. As reported by the South China Morning Post, Facebook has over five million users in Hong Kong. By rolling out the chatbot on a commonly used platform, Citi is going where its users are.

The US and China are leagues ahead of any other country when it comes to AI technology. And it's because they know how to prioritize their own programs. Rather than waste time discussing the dangers of AI with the UN, or crafting global policy, the two countries have become AI nationalists. Artificial intelligence research was a dying field just a decade ago. Now, AI companies like Google and Baidu are among the richest in the world.

SYDNEY – Australia will invest 7 billion Australian dollars ($5.2 billion) to develop and buy high-tech U.S. drones for joint military operations and to monitor waters including the South China Sea, it said Tuesday. Canberra has been embarking on its largest peacetime naval investment through a massive shipbuilding strategy that includes new submarines, offshore patrol vessels and frigates to shore up its defense capabilities. As part of this, the government will spend AU$1.4 billion to buy the first of six MQ-4C Triton maritime surveillance drones, with the aircraft to enter service from mid-2023, complementing seven P-8A Poseidon planes currently in use. "Together these aircraft will significantly enhance our anti-submarine warfare and maritime strike capability, as well as our search and rescue capability," Prime Minister Malcolm Turnbull said in a statement. "This investment will protect our borders and make our region more secure."

It's been just over 30 years since the last major overall of the U.S. tax code. In that time the world has been transformed - the Soviet Union collapsed, the Berlin Wall fell, dot coms boomed and busted, terrorism struck and launched the U.S. into the longest war in its history. A financial crisis shook the country and the world to its knees, and the rise of big data, artificial intelligence, genomics, new materials, cloud computing, blockchain, the sharing and gig economies, and many other new, advanced technologies and business models signaled the start of the Fourth Industrial Revolution. In that time, tax law and accounting practices have struggled to keep pace with innovations, sometimes leading to a wild West free for all for businesses and consumers, some of who managed to profit while others lost or were swindled of their life savings. Even today tax laws still fail to address the issues that the Age of Computer and the Internet brought, such as the internet sales tax, which may be heading to the US Supreme Court in the next year or so. As President Trump and the U.S. Congress are poised to pass and celebrate the passage of reforming the U.S. tax code for the first time in 30 years, missing from the discussion by both parties is how tax law and accounting principles should be altered to account for the realities of how AI, automation and the Fourth Industrial Revolution are reshaping businesses and the labor markets today and will continue to transform them in the years to come.

This photo taken on February 5, 2018 shows a police officer wearing a pair of smartglasses with a facial recognition system at Zhengzhou East Railway Station in Zhengzhou in China's central Henan province. Chinese police are sporting high-tech sunglasses that can spot suspects in a crowded train station, the newest use of facial recognition that has drawn concerns among human rights groups. We seem to be heading into a future where facial recognition technologies are going to be part of everyday life. Cities all over the world are now bristling with cameras, and in the case of China it is impossible to avoid being monitored either by CCTV or even by police wearing special glasses and then logged onto a database that checks on your habits, your social credit and even who your friends are. At the same time, cameras and facial recognition are increasingly being used in public and private buildings.

Many modern machine learning models are trained to achieve zero or near-zero training error in order to obtain near-optimal (but non-zero) test error. This phenomenon of strong generalization performance for "overfitted" / interpolated classifiers appears to be ubiquitous in high-dimensional data, having been observed in deep networks, kernel machines, boosting and random forests. Their performance is robust even when the data contain large amounts of label noise. Very little theory is available to explain these observations. The vast majority of theoretical analyses of generalization allows for interpolation only when there is little or no label noise. This paper takes a step toward a theoretical foundation for interpolated classifiers by analyzing local interpolating schemes, including geometric simplicial interpolation algorithm and weighted $k$-nearest neighbor schemes. Consistency or near-consistency is proved for these schemes in classification and regression problems. These schemes have an inductive bias that benefits from higher dimension, a kind of "blessing of dimensionality". Finally, connections to kernel machines, random forests, and adversarial examples in the interpolated regime are discussed.

A long-standing problem in the theory of stochastic gradient descent (SGD) is to prove that its without-replacement version RandomShuffle converges faster than the usual with-replacement version. We present the first (to our knowledge) non-asymptotic solution to this problem, which shows that after a "reasonable" number of epochs RandomShuffle indeed converges faster than SGD. Specifically, we prove that under strong convexity and second-order smoothness, the sequence generated by RandomShuffle converges to the optimal solution at the rate O(1/T^2 + n^3/T^3), where n is the number of components in the objective, and T is the total number of iterations. This result shows that after a reasonable number of epochs RandomShuffle is strictly better than SGD (which converges as O(1/T)). The key step toward showing this better dependence on T is the introduction of n into the bound; and as our analysis will show, in general a dependence on n is unavoidable without further changes to the algorithm. We show that for sparse data RandomShuffle has the rate O(1/T^2), again strictly better than SGD. Furthermore, we discuss extensions to nonconvex gradient dominated functions, as well as non-strongly convex settings.

We consider the following conditional linear regression problem: the task is to identify both (i) a $k$-DNF condition $c$ and (ii) a linear rule $f$ such that the probability of $c$ is (approximately) at least some given bound $\mu$, and $f$ minimizes the $\ell_p$ loss of predicting the target $z$ in the distribution of examples conditioned on $c$. Thus, the task is to identify a portion of the distribution on which a linear rule can provide a good fit. Algorithms for this task are useful in cases where simple, learnable rules only accurately model portions of the distribution. The prior state-of-the-art for such algorithms could only guarantee finding a condition of probability $\Omega(\mu/n^k)$ when a condition of probability $\mu$ exists, and achieved an $O(n^k)$-approximation to the target loss, where $n$ is the number of Boolean attributes. Here, we give efficient algorithms for solving this task with a condition $c$ that nearly matches the probability of the ideal condition, while also improving the approximation to the target loss. We also give an algorithm for finding a $k$-DNF reference class for prediction at a given query point, that obtains a sparse regression fit that has loss within $O(n^k)$ of optimal among all sparse regression parameters and sufficiently large $k$-DNF reference classes containing the query point.

Classic Topic Models are built under the Bag Of Words assumption, in which word position is ignored for simplicity. Besides, symmetric priors are typically used in most applications. In order to easily learn topics with different properties among the same corpus, we propose a new line of work in which the paragraph structure is exploited. Our proposal is based on the following assumption: in many text document corpora there are formal constraints shared across all the collection, e.g. sections. When this assumption is satisfied, some paragraphs may be related to general concepts shared by all documents in the corpus, while others would contain the genuine description of documents. Assuming each paragraph can be semantically more general, specific, or hybrid, we look for ways to measure this, transferring this distinction to topics and being able to learn what we call specific and general topics. Experiments show that this is a proper methodology to highlight certain paragraphs in structured documents at the same time we learn interesting and more diverse topics.

We provide tight finite-time convergence bounds for gradient descent and stochastic gradient descent on quadratic functions, when the gradients are delayed and reflect iterates from $\tau$ rounds ago. First, we show that without stochastic noise, delays strongly affect the attainable optimization error: In fact, the error can be as bad as non-delayed gradient descent ran on only $1/\tau$ of the gradients. In sharp contrast, we quantify how stochastic noise makes the effect of delays negligible, improving on previous work which only showed this phenomenon asymptotically or for much smaller delays. Also, in the context of distributed optimization, the results indicate that the performance of gradient descent with delays is competitive with synchronous approaches such as mini-batching. Our results are based on a novel technique for analyzing convergence of optimization algorithms using generating functions.