This article originally appeared in the March 2017 issue of IEEE Robotics & Automation Magazine. We thank RAM and the authors for giving us permission to reproduce it here. Algorithms with learning abilities collect personal data that are then used without users' consent and even without their knowledge; autonomous weapons are under discussion in the United Nations; robots stimulating emotions are deployed with vulnerable people; research projects are funded to develop humanoid robots; and artificial intelligence-based systems are used to evaluate people. One can consider these examples of AI and autonomous systems (AS) as great achievements or claim that they are endangering human freedom and dignity. We need to make sure that these technologies are aligned to humans in terms of our moral values and ethical principles to fully benefit from the potential of them.

This article presents a logical formalization of the emotional appraisal theory, i.e., it formalizes the cognitive process of evaluation that elicits an emotion. This formalization is psychologically grounded on the OCC cognitive model of emotions. More specifically, we are interested in event-based emotions, i.e., emotions that are elicited by the evaluation of the consequences of an event that either happened or will happen. The formal modelling presented here is based on the AfPL Probabilistic Logic, a BDI-like probabilistic modal logic, which allows our model to verify whether the variables that determine the elicitation of emotions achieved the necessary threshold or not. The proposed logical formalization aims at addressing how the emotions are elicited by the agent cognitive mental states (desires, beliefs and intentions), and how to represent the intensity of the emotions. These are important initial points in the investigation of the dynamic interaction among emotions and other mental states.

In a previous story, I wrote about how a Game Theoretic approach was influencing developments in the Deep Learning field. In this story, I now write about DeepMind's latest foray into this exciting area. Yesterday, February 19th 2017), DeepMind presents their latest research on this subject titled "Understanding Agent Cooperation". The gist of the research is that, they employed Deep Reinforcement Learning networks in two game environments to study their behavior. The motivation is to study multi-agent systems to better understand and control these kinds of systems. In a previous story (see: "Five Capability Levels of Deep Learning", I laid out a road map as to how Deep Learning will evolve in even greater capabilities.

Inverse reinforcement learning (IRL) has become a useful tool for learning behavioral models from demonstration data. However, IRL remains mostly unexplored for multi-agent systems. In this paper, we show how the principle of IRL can be extended to homogeneous large-scale problems, inspired by the collective swarming behavior of natural systems. In particular, we make the following contributions to the field: 1) We introduce the swarMDP framework, a sub-class of decentralized partially observable Markov decision processes endowed with a swarm characterization. 2) Exploiting the inherent homogeneity of this framework, we reduce the resulting multi-agent IRL problem to a single-agent one by proving that the agent-specific value functions in this model coincide. 3) To solve the corresponding control problem, we propose a novel heterogeneous learning scheme that is particularly tailored to the swarm setting. Results on two example systems demonstrate that our framework is able to produce meaningful local reward models from which we can replicate the observed global system dynamics.

New research from OpenAI and UC Berkeley has created A.I. agents that can form and use their own new language, without instruction, whenever they need to. The languages are systematic and roughly grammatical, and even include aspects of non-verbal communication like body language! It all makes for an incredible glimpse into how (and why) language may have arisen during biological evolution, and it shows the nuanced insight we can derive from modern learning agents. Like so many studies that set out to elicit a specific A.I. behavior, this one began by creating a rough metaphor for real life. The experiment sets its A.I. agents in a simulated physical world containing landmarks at fixed positions, and then gives them the ability to roam freely within this two-dimensional space. The agents were then given a goal, usually to send another agent to a specific place in the world, and a set of nonsense symbols each could "say" aloud so the others could "hear" it.

Which authors of this paper are endorsers? Disable MathJax (What is MathJax?)

Scientists are conducting a massive computer simulation to work out how New York would respond to a nuclear attack in the heart of Manhattan. The three-year, $450,000 project will simulate two nuclear detonations and their effects on up to 20 million virtual'agents' each representing civilian, first responder or other official over the course of 30 days. But first they need to input data - a lot of data, taken from disaster reports across the US - to figure out how individuals really react to catastrophe. 'Computational social science is not experimental.' Professor William Kennedy of Virginia's George Mason University told The Atlantic.

During the past decade, several areas of speech and language understanding have witnessed substantial breakthroughs from the use of data-driven models. In the area of dialogue systems, the trend is less obvious, and most practical systems are still built through significant engineering and expert knowledge. Nevertheless, several recent results suggest that data-driven approaches are feasible and quite promising. To facilitate research in this area, we have carried out a wide survey of publicly available datasets suitable for data-driven learning of dialogue systems. We discuss important characteristics of these datasets, how they can be used to learn diverse dialogue strategies, and their other potential uses. We also examine methods for transfer learning between datasets and the use of external knowledge. Finally, we discuss appropriate choice of evaluation metrics for the learning objective.

Sharon Glotzer has made a number of career-shifting discoveries, each one the kind "that completely changes the way you look at the world," she said, "and causes you to say, 'Wow, I need to follow this.'" A theoretical soft condensed matter physicist by training who now heads a thriving 33-person research group spanning three departments at the University of Michigan in Ann Arbor, Glotzer uses computer simulations to study emergence--the phenomenon whereby simple objects give rise to surprising collective behaviors. "When flocks of starlings make these incredible patterns in the sky that look like they're not even real, the way they're changing constantly--people have been seeing those patterns since people were on the planet," she said. "But only recently have scientists started to ask the question, how do they do that? How are the birds communicating so that it seems like they're all following a blueprint?"

OpenAI research has a short introduction on their newest research "Learning to Communicate". There are many trends that I watch for in the field of Deep Learning. Two trends that are related and I believe going to be very promising areas are language learning and multi-agent communication. If you have not been watching, this week has had a tremendous release of papers involving the former and culminating with OpenAI's post, stitching it all together! Let me explain though what transpired in this amazing week.