Marketers have been talking about artificial intelligence (AI) marketing for some time now. But many are still wondering, what exactly can AI marketing do for us right now? To find out, Econsultancy recently held a gathering of over 400 marketers for Digital Outlook 2018 in Singapore and invited subject matter experts to speak on AI marketing and other topics trending in digital. In his talk, Chandra Kumar, CEO WiselyWise, stated that, despite many setbacks over its long history, AI marketing may now be ready for mass adoption. He proceeded to describe four concrete things that marketers can use AI for right now.

We investigate the effects of initialization and architecture on the start of training in deep ReLU nets. We identify two common failure modes for early training in which the mean and variance of activations are poorly behaved. For each failure mode, we give a rigorous proof of when it occurs at initialization and how to avoid it. The first failure mode, exploding/vanishing mean activation length, can be avoided by initializing weights from a symmetric distribution with variance 2/fan-in. The second failure mode, exponentially large variance of activation length, can be avoided by keeping constant the sum of the reciprocals of layer widths. We demonstrate empirically the effectiveness of our theoretical results in predicting when networks are able to start training. In particular, we note that many popular initializations fail our criteria, whereas correct initialization and architecture allows much deeper networks to be trained.

Network quantization is an effective solution to compress deep neural networks for practical usage. Existing network quantization methods cannot sufficiently exploit the depth information to generate low-bit compressed network. In this paper, we propose two novel network quantization approaches, single-level network quantization (SLQ) for high-bit quantization and multi-level network quantization (MLQ) for extremely low-bit quantization (ternary).We are the first to consider the network quantization from both width and depth level. In the width level, parameters are divided into two parts: one for quantization and the other for re-training to eliminate the quantization loss. SLQ leverages the distribution of the parameters to improve the width level. In the depth level, we introduce incremental layer compensation to quantize layers iteratively which decreases the quantization loss in each iteration. The proposed approaches are validated with extensive experiments based on the state-of-the-art neural networks including AlexNet, VGG-16, GoogleNet and ResNet-18. Both SLQ and MLQ achieve impressive results.

In this paper, we consider the convergence of an abstract inexact nonconvex and nonsmooth algorithm. We promise a pseudo sufficient descent condition and a pseudo relative error condition, which both are related to an auxiliary sequence, for the algorithm; and a continuity condition is assumed to hold. In fact, a wide of classical inexact nonconvex and nonsmooth algorithms allow these three conditions. Under the finite energy assumption on the auxiliary sequence, we prove the sequence generated by the general algorithm converges to a critical point of the objective function if being assumed Kurdyka- Lojasiewicz property. The core of the proofs lies on building a new Lyapunov function, whose successive difference provides a bound for the successive difference of the points generated by the algorithm. And then, we apply our findings to several classical nonconvex iterative algorithms and derive corresponding convergence results.

Of all the purported uses of Artificial Intelligence (AI), it would be hard to find one more controversial than its possible use for military purposes. In popular consciousness, the idea of military AI immediately brings to mind the notion of autonomous weapon systems or "killer robots", machines that can independently target and kill humans. The possible presence of such systems on battlefields has sparked a welcome international debate on the legality and morality of using these weapon systems. The controversies surrounding autonomous weapons, however, must not obscure the fact that like most technologies, AI has a number of non-lethal uses for militaries across the world, and especially for the Indian military. These are, on the whole, not as controversial as the use of AI for autonomous weapons, and, in fact, are far more practicable at the moment, with clear demonstrable benefits.

The tech industry is desperate for AI talent, with some companies forking over $1 million or more to secure the services of those who can code their way to machine intelligence. The problem, however, is that hardly anyone is qualified to evaluate the relative merits of any particular job candidate. As such, employers are almost certainly overpaying for AI talent, and may not be getting much in the way of talent, to boot. "Google is paying a million dollars for these superstars," declared Kai-Fu Lee, Google's former China chief, at the World Economic Forum. "You may not need someone that high, but you've got to break the scale for at least one person," Lee said.

Last month Airbus released a video of the first successful test flight of its electric vertical take-off and landing (eVTOL) autonomous drone. Although it only hovered in the air for 53 seconds, the fact that its eight rotors were powered entirely by electricity was a landmark for the manufacturer of gas-guzzling planes. The goal is that the technology could be used for airborne travel in congested cities. "Our goal is to democratise personal flight by leveraging the latest technologies such as electric propulsion, energy storage and machine vision," blogged Zach Lovering, Vahana project executive. Chinese drone manufacturer Ehang is considerably more advanced than Airbus. In February it flew 40 journalists and local dignitaries on trips of up to 15km in Guangzhou, southern China, reaching top speeds of 80mph.

Elon Musk, the billionaire tech entrepreneur, has also voiced fears over the potential threat AI technology poses. "If you're not concerned about AI safety, you should be. Vastly more risk than North Korea," he said last year. However Mr Schmidt argued that the benefits AI brings to technology and medical advances outweigh concerns about the negative effects - and he insisted "humans will remain in charge of [AI] for the rest of time". He said: "Everyone immediately then wants to talk about all the movie-inspired death scenarios, and I can confidently predict to you that they are one to two decades away. "So let's worry about them, but let's worry about them in a while." He went on to say that the technology had major flaws and would always be within the control of humans. He said: "I want to remind everyone these technologies [AI] have serious errors in them and they should not be used with life-critical decisions.

China's leading technology companies are on fire, heavily investing in artificial intelligence and building true global presences. McKinsey recently reported that academic and research institutions in the country publish more cited research papers than the US, UK, or any other global leader in AI, producing nearly 10,000 papers in 2015 alone. Backed by strong government mandates and billions of dollars of both private and public investments, China is challenging the US for position of global AI leader. Fearful of competition, the US government is considering placing restrictions on Chinese investments in AI and technology in the United States. In many sectors, such as healthcare, China may already be ahead of America in applying AI to critical public issues.

Toyota has invested ¥300 billion ($2.8 billion) in a new Tokyo-based company that will build software for self-driving cars, according to a report in the Wall Street Journal on Friday. The Toyota Research Institute-Advanced Development (or TRI-AD) will set out with 300 employees but the Japanese automaker hopes that number will grow to about 1,000 as the company takes off. Toyota has an ambitious deadline of 2020 for testing its autonomous, electric vehicles, which is part of the company's larger goal of commercially expanding the availability and capabilities of self-driving cars. When Toyota first announced its 2020 goal, the company teased that the autonomous vehicles in development might come with AI functionalities that could facilitate conversations between vehicles and passengers. Toyota's main goal with TRI-AD is ensuring the development of reliable software for the cars.