Composition of elementary skills into complex behaviors to solve challenging problems is one of the key elements toward building intelligent machines. To date, there has been plenty of work on learning new policies or skills but almost no focus on composing them to perform complex decision-making. In this paper, we propose a policy ensemble composition framework that takes the robot's primitive policies and learns to compose them concurrently or sequentially through reinforcement learning. We evaluate our method in problems where traditional approaches either fail or exhibit high sample complexity to find a solution. We show that our method not only solves the problems that require both task and motion planning but also exhibits high data efficiency, which is currently one of the main limitations of reinforcement learning.

In order to be useful in the real world, AI agents need to plan and act in the presence of others, who may include adversarial and cooperative entities. In this paper, we consider the problem where an autonomous agent needs to act in a manner that clarifies its objectives to cooperative entities while preventing adversarial entities from inferring those objectives. We show that this problem is solvable when cooperative entities and adversarial entities use different types of sensors and/or prior knowledge. We develop two new solution approaches for computing such plans. One approach provides an optimal solution to the problem by using an IP solver to provide maximum obfuscation for adversarial entities while providing maximum legibility for cooperative entities in the environment, whereas the other approach provides a satisficing solution using heuristic-guided forward search to achieve preset levels of obfuscation and legibility for adversarial and cooperative entities respectively. We show the feasibility and utility of our algorithms through extensive empirical evaluation on problems derived from planning benchmarks.

Deep reinforcement learning (RL) policies are known to be vulnerable to adversarial perturbations to their observations, similar to adversarial examples for classifiers. However, an attacker is not usually able to directly modify another agent's observations. This might lead one to wonder: is it possible to attack an RL agent simply by choosing an adversarial policy acting in a multi-agent environment so as to create natural observations that are adversarial? We demonstrate the existence of adversarial policies in zero-sum games between simulated humanoid robots with proprioceptive observations, against state-of-the-art victims trained via self-play to be robust to opponents. The adversarial policies reliably win against the victims but generate seemingly random and uncoordinated behavior. We find that these policies are more successful in high-dimensional environments, and induce substantially different activations in the victim policy network than when the victim plays against a normal opponent.

Deepfakes are video manipulations that can make people say seemingly strange things. Barack Obama and Nicolas Cage have been featured in these videos. We all know about Mona Lisa's smile. Now, you can watch her talk. A research paper from experts at Samsung's AI Center in Moscow and the Skolkovo Institute of Science and Technology shows how fake videos can be created with a single image, including the classic artwork.

Amazon is offering up $25 gift cards in exchange for 3D scans of your body. The internet giant is currently conducting a study at its New York office as part of Amazon Body Labs that seeks to'learn about diversity among body shapes,' according to a listing, which was first spotted by Mashable. Participants who set up a 30-minute appointment will be asked to take a survey and agree to have 3D scans, photos and videos taken of them. The move comes as Amazon has faced privacy concerns around its collection of data from Echo devices, as well as the deployment of its controversial facial recognition software. Amazon is offering up $25 gift cards in exchange for 3D scans of your body.

The deployment of machine learning models is the process for making your models available in production environments, where they can provide predictions to other software systems. It is only once models are deployed to production that they start adding value, making deployment a crucial step. However, there is complexity in the deployment of machine learning models. This post aims to at the very least make you aware of where this complexity comes from, and I'm also hoping it will provide you with useful tools and heuristics to combat this complexity. If it's code, step-by-step tutorials and example projects you are looking for, you might be interested in the Udemy Course "Deployment of Machine Learning Models".

"AI is being fed directly into the bloodstream of society, and in many cases without sufficient checks and balances," says Kate Crawford, a professor and cofounder of New York University's AI Now, the world's first academic research institute dedicated to the social impact of artificial intelligence. Last year, Crawford partnered with data-viz guru Vladan Joler to create "Anatomy of an AI System," a map and research paper demonstrating the real-world consequences of developing and manufacturing the Amazon Echo. The paper highlights the radical differences in income distribution between Amazon executives and the workers who enable its vast infrastructure, as well as its devastating environmental impacts. The project has been exhibited at museums around the world, and Crawford has presented it to leaders in France, Germany, Spain, and Argentina.

This week Amazon CEO Jeff Bezos got the tech sector's attention with emerging reports of his "fascination" with the rapidly developing world of autonomous autos. "If you think about the auto industry right now, there's so many things going on with Uber-ization, electrification, the connected car -- so it's a fascinating industry," Bezos said. "It's going to be something very interesting to watch and participate in, and I'm very excited about that whole industry." Amazon has made some sizable investments to accompany that interest -- most notably in automation and electrification start-up Rivian and self-driving startup Aurora. And fascination aside, Amazon has a race for the consumer's whole paycheck to vie in with Walmart -- and there is little doubt that auto automation plays like Rivian and Aurora could put some octane, so to speak, behind that effort.

A new contract with the Massachusetts Institute of Technology (MIT) will bring airmen from across Air Force career fields to work with researchers on artificial intelligence technology. The project will focus on research in AI projects including decision support, maintenance and logistics, talent management, medical readiness, situational awareness, business operations and disaster relief, according to a news release. The effort is part of the service's science and technology strategy. Similar partnerships around the U.S. focus on other innovations.