When building machine learning models, we want to keep error as low as possible. Two major sources of error are bias and variance. If we managed to reduce these two, then we could build more accurate models. But how do we diagnose bias and variance in the first place? And what actions should we take once we've detected something?

Advances in superconducting materials and more efficient passive photovoltaic or other renewable energy sources coupled with real world locomotion will limit the ability to control AI in its myriad potential deployments. Self-driving cars are a self-contained, albeit enormously complex task, but represent just a sliver of the overall movement negotiation handled by a human under self-power or controlling another vehicle. The negotiation of movement into learned models is a function of morphology. Much as Darwin's theory of natural selection applies, this is a basic outgrowth of a competitive human environment where an AI would be designed to be the "best" at what is supposed to do. This doesn't mean explicit kill logic will be present.

As I sit down in Nissan's simulator, I prepare myself for the fact that a cohort of researchers could scrutinize my skills as a wheelman with more rigor than the most aggravating backseat driver. And, I accept that this process involves wearing what looks like a too-small, sideways bicycle helmet, which holds 11 electrodes poking through my hair. "For each corner, there'll be an evaluation of your driving smoothness," says Lucian Gheorghe, the Nissan researcher in charge of this rig. Gheorghe is interested in motor related potentials, a specific pattern of activity the brain creates as it prepares to move a limb. It takes half a second for the body to translate that signal to the wave of an arm or kick of a leg, and Nissan wants to exploit the gap.

Honda introduced its new 3E (Empower, Experience, Empathy) Robotics Concept at CES 2018, demonstrating a range of experimental technologies engineered to understand people's needs and make their lives better. Through a suite of robotic concepts expressing a variety of functions and designs, Honda shared its vision of a society where robotics and AI can assist people in many situations such as disaster recovery, recreation and learning from human interaction to become more helpful and empathetic. Honda's 3E Robotics Concept is part of the company's core areas of focus at CES 2018: robotics, mobility, and energy management. Honda intends to pursue these areas through its "open innovation" approach, developing technology themes that foster collaboration with partners who share Honda's vision. Rather than stand-alone devices that work independently, Honda envisions robotics as multiple devices that work together as a system, enabling people to expand their life's potential.

Though not an alternative to human knowledge and ingenuity, AI is considered a supporting tool to help humans. Though Al presently has a tough time completing different tasks involving common sense in the real world, it is able to process large amounts of data faster compared to a human brain.

There is great enthusiasm for the smart city concept. Integration of autonomous vehicles, drones and networked communications are expected to manage congestion, lead to fewer accidents, reduce pollution and enhance quality of life. The smart city was a major theme at the 2018 Consumer Electronics Show (#CES2018), hosted by the Consumer Technology Association. Will smart cities be vibrant bastions of competitive private free enterprise and innovative new networks of communication that simultaneously respect individuals' privacy? Or are planners on a path to setting up mega public utilities and administered cartelization, and compulsory information collection?

In this paper, an interference-aware path planning scheme for a network of cellular-connected unmanned aerial vehicles (UAVs) is proposed. In particular, each UAV aims at achieving a tradeoff between maximizing energy efficiency and minimizing both wireless latency and the interference level caused on the ground network along its path. The problem is cast as a dynamic game among UAVs. To solve this game, a deep reinforcement learning algorithm, based on echo state network (ESN) cells, is proposed. The introduced deep ESN architecture is trained to allow each UAV to map each observation of the network state to an action, with the goal of minimizing a sequence of time-dependent utility functions. Each UAV uses ESN to learn its optimal path, transmission power level, and cell association vector at different locations along its path. The proposed algorithm is shown to reach a subgame perfect Nash equilibrium (SPNE) upon convergence. Moreover, an upper and lower bound for the altitude of the UAVs is derived thus reducing the computational complexity of the proposed algorithm. Simulation results show that the proposed scheme achieves better wireless latency per UAV and rate per ground user (UE) while requiring a number of steps that is comparable to a heuristic baseline that considers moving via the shortest distance towards the corresponding destinations. The results also show that the optimal altitude of the UAVs varies based on the ground network density and the UE data rate requirements and plays a vital role in minimizing the interference level on the ground UEs as well as the wireless transmission delay of the UAV.

Stochastic gradient descent (SGD) is widely believed to perform implicit regularization when used to train deep neural networks, but the precise manner in which this occurs has thus far been elusive. We prove that SGD minimizes an average potential over the posterior distribution of weights along with an entropic regularization term. This potential is however not the original loss function in general. So SGD does perform variational inference, but for a different loss than the one used to compute the gradients. Even more surprisingly, SGD does not even converge in the classical sense: we show that the most likely trajectories of SGD for deep networks do not behave like Brownian motion around critical points. Instead, they resemble closed loops with deterministic components. We prove that such "out-of-equilibrium" behavior is a consequence of highly non-isotropic gradient noise in SGD; the covariance matrix of mini-batch gradients for deep networks has a rank as small as 1% of its dimension. We provide extensive empirical validation of these claims, proven in the appendix.

Apple's spaceship is getting its finishing touches. The latest drone video of the $5bn HQ have revealed the final landscaping touches being made to the huge campus. It shows the landscaping that has transformed the giant building site into a lush green park. The drone footage reveals the incredible landscaping inside the giant ring to turn it into a park. Apple Park contains over 9,000 native and drought-resistant trees, and is powered by 100 percent renewable energy.

We are living in interesting times. Multiple technologies, improving exponentially, are converging. I have been chronicling this convergence for several years in my newsletter, Exponential View. As Bill Gates said, "Most people overestimate what they can do in one year and underestimate what they can do in ten years." Likewise, most annual predictions overestimate what can occur in a year, and underestimate the power of the trend over time.