Montezuma's Revenge is one of the most challenging Atari games An artificial intelligence that can remember its previous successes and use them to create new strategies has achieved record high scores on some of the hardest video games on classic Atari consoles. Many AI systems use reinforcement learning, in which an algorithm is given positive or negative feedback on its progress towards a particular goal after each step it takes, encouraging it towards a particular solution. This technique was used by AI firm DeepMind to train AlphaGo, which beat a world champion Go player in 2016. Adrien Ecoffet at Uber AI Labs and OpenAI in California and his colleagues hypothesised that such algorithms often stumble upon encouraging avenues but then jump to another area in the hunt for something more promising, leaving better solutions overlooked. "What do you do when you don't know anything about your task?" says Ecoffet. "If you just wave your arms around, it's unlikely that you're ever going to make a coffee."

GPT-3 is the world's most sophisticated natural language technology. Discover how companies are implementing the OpenAI GPT-3 API to power new use cases.

AI systems, on the other hand, are built to do only one of these things at a time. Computer-vision and audio-recognition algorithms can sense things but cannot use language to describe them. A natural- language model can manipulate words, but the words are detached from any sensory reality. If senses and language were combined to give an AI a more human-like way to gather and process new information, could it finally develop something like an understanding of the world? The hope is that these "multimodal" systems, with access to both the sensory and linguistic "modes" of human intelligence, should give rise to a more robust kind of AI that can adapt more easily to new situations or problems.

This tutorial provides a simple introduction to using multi-agent reinforcement learning, assuming a little experience in machine learning and knowledge of Python. Reinforcement stems from using machine learning to optimally control an agent in an environment. It works by learning a policy, a function that maps an observation obtained from its environment to an action. Policy functions are typically deep neural networks, which gives rise to the name "deep reinforcement learning." The goal of reinforcement learning is to learn an optimal policy, a policy that achieves the maximum expected reward from the environment when acting.

Arduino Deep Learning From Ground Up Build Artificial Intelligence Sketch from Scratch on Arduino What you'll learn Welcome to the Arduino Deep Learning From Ground Up . We are going to embark on a very exciting journey together. We are going to learn how to build deep neural networks from scratch on our Arduino. We shall begin by learning the basics of deep learning with practical code showing each of the basic building blocks that end up making a giant deep neural network. As we begin to deal with large datasets we shall start training our neural networks on our computers and then deploying the the trained models on our microcontrollers.

Artificial intelligence allows machines to understand and achieve specific goals. AI includes machine learning via deep learning.

The vision of Robots and Humans working together has been popularized by numerous Hollywood movies, comic books and the media for many decades now. Success stories about the application of automation in the manufacturing industry date back to the early 1980’s, when welding robots demonstrated efficiency and resilience to deliver on time and increased the overall supply. And ever since, robots have been used for a variety of applications such as painting, assembly, disassembly, pick and place for printed circuit boards, material handling, product inspection, and testing, all accomplished with high speed and precision. In the past few years, however, the world is seeing a significant shift towards AI-enabled robots that are changing the game by bringing various levels of autonomy into the picture. If we consider the example of the labor-intensive task of warehouse piece picking,  AI-enabled robots are learning to handle millions of objects with minimum help from humans. What once required individual item registration and programming of robots, can now be accomplished with self-directing or autonomous robots using deep learning algorithms. This in turn, is helping industries cut costs and improve profits in the long run. Further, with the dawn of COVID-19, the way we patronize restaurants and shops and see our doctors has changed drastically. From collaborative robots (co-robots) to robots with high levels of autonomy, the trend continues towards the application of Autonomous Mobile Robots (AMR) in every industry and retail setting; more so as societies are restructuring places of work and pleasure to minimize human contact. The accelerated adoption of AMRs is being seen in many frontline roles from spraying disinfectants to delivering & serving food to customers, in outdoor and indoor urban hotspots. A key advantage of using AMRs is that they are able to reliably do these repetitive physical tasks when many workers aren’t safely able to or willing to set foot in these areas. As AMRs are becoming more and more competent, a few popular ones that are gaining traction during this pandemic are discussed below.   TELEPRESENCE ROBOT Telepresence Robots or Virtual Assistance Robots, now more than ever, are revolutionizing the way we work or learn remotely. They enable telecommuters, doctors, remote workers, and students to feel more connected to their colleagues by giving them a physical presence where they can’t be in person.   Image Credit: Using robots to enable patients to be inspected remotely In recent times, Telepresence robots are being put to use to enable interactions with COVID-19 patients in isolation wards. These semi-autonomous robots could be teleoperated over the Wi-Fi, to visit the patients and provide a live video link to their loved ones or to deliver their food and medication.   DISINFECTANT ROBOT  In the post pandemic world, disinfecting shared public places has been a tricky problem to solve. A non-intrusive solution widely adopted in recent times is UV-C light, but due to its adverse effects on human skin, trolleying it would look somewhat like the image below.   Image credit: UV-C based disinfection trolley by rapid cleaning of hospital environment, helping in the fight against COVID-19 Enter AMR, mounted with a UV-C lamp as a payload, which would traverse autonomously in a designated area.                                                                                     Image Credit: Weston Robot   This is accomplished by using the mobile app where the user creates a map of the area and chooses the waypoints for the robot to traverse. The robot then follows the waypoints and stops at each waypoint, till the disinfection job is done. Also, it returns to the charging station once the battery is below a threshold eliminating the need for human intervention. A swarm of these robots could be used in and around the areas where the most interactions occur such as hospitals, railway stations, airports, etc. An intrusive solution would be to air-blast disinfectant liquids similar to the ones used in agricultural spraying, to increase the chlorine content in the air and reduce the possibility of aerosol transmission.   Image Credit: Robots deployed to disinfect open spaces This fully autonomous solution is basically an AMR, mounted with a sprayer mechanism, which can navigate to the decontamination area using pre-built maps giving  the technical staff less exposure to these highly concentrated chemicals while working in an entirely safe and risk-free environment. The range of its spray devices can reach up-to 30 feet.   LAST-MILE DELIVERY ROBOTS As door delivery businesses saw a huge increase in demand during the pandemic, Last-Mile delivery robots found application as a reliable and safe contactless delivery system. These robots are an alternative to human food delivery drivers from companies like Uber Eats and DoorDash, which perform tasks that a person cannot do safely. These companies have created and deployed cool new robots with the intelligence to navigate city streets to deliver orders from selected restaurants to the customer location using a mobile app, while avoiding dynamic obstacles like pedestrians.   Image Credit: Autonomous delivery robot   The robot’s body is equipped with a storage bin with a locking compartment where the restaurant stores the delivery package and the bin only unlocks upon authentication by the customer at her location during delivery. Typically, these robots are equipped with cameras and computer vision driven by machine learning. They can detect and classify what they see, and tell the difference between a car, a person, or a wall. While these robots are definitely cool tech; procuring, running and maintaining fleets of robots can be prohibitively expensive. To circumvent this, the solution that small restaurant owners are turning to is the clever ‘Robot-as-a-Service’(RaaS) business model for food delivery, which is becoming more crucial as Covid-19 reshapes the gig economy.   ROBOT AS A SERVICE Many are now familiar with the concept of Software as a Service (SaaS) or Big Data as a Service (BDaaS) or Platform as a Service (PaaS) where the intent is to democratize technology while lowering the barrier to entry for businesses, large and small alike. One of the new areas, this philosophy is becoming more prevalent, is “Robot as a Service” (RaaS), a cloud-based “robotic rental” solution used for both B2B or B2C businesses. RaaS takes the capabilities of robotics and removes the upfront cost of robot installation with large amounts of computing power, utilities, and knowledge.  And so, small- and medium-sized businesses are increasingly experimenting with RaaS because of its flexibility, scalability, and lower cost of entry compared to traditional robotics programs. For example, finding sufficient  numbers of workers in a warehouse, for online retailers during seasonal surges is quite the challenge. With the RaaS model, these seasonal labor shortcomings can be mitigated without investing in equipment that won’t be used in slower periods while still being able to quickly scale up to meet the high demand. RaaS might be the answer for businesses, trying to figure out how to improve productivity or reduce risk, but always thought, robots were out of their price range.   ROBOTICS AT IGNITARIUM Robotics and in particular Autonomous Mobile Robots is an area where Ignitarium has been developing technology solutions for the warehouse use cases. Robotics as a theme started as a R&D thread to leverage the existing skills in computer vision and AI/ML. In the past year, we have demonstrated various use cases on sensor fusion, integration of Lidars & Radars, visual odometry using RGB-D Cameras, path planning, obstacle avoidance, integrating with deep learning modules for object Classification / Detection / Tracking. Our vision is to create a software package which is hardware agnostic, for use  across hardware platforms with minimal customization.     CONCLUSION The COVID-19 pandemic has accelerated the adoption of Autonomous Mobile Robots (AMRs) such as Telepresence, Disinfectant, Last-mile delivery robots. Roboticists are also seeing adaptation of AMRs to new niches and are exploring new avenues. Although the robotics wave is changing the way services and products are sold, there are still challenges to overcome like the amount of customization required, both in hardware and software for the robots to adapt to customer specific needs. Regardless of the hurdles, RaaS will be the inevitable solution many organizations seek either with hardware & software or software-only flavors.

If there is any doubt in your mind that are not headed to a future where mind-machine meld is going to be the new norm, just look at Elon Musk's Neuralink's BCI. The animal trials are already underway, as claimed by Musk, a monkey with a wireless implant in his skull with tiny wires can play video games with his mind. Although designed to cure a wide variety of diseases, the experiment aligns with Musk's long-term vision of coming up with a brain-computer interface that is able to compete with increasingly powerful AIs. However, Neuralink's proposed device is an invasive one that requires fine threads that need to be implanted in the brain. And as if these invasive devices were not scary enough for a person like me, new breakthroughs in neuroscience and artificial intelligence might infiltrate our emotions -- the last bastion of personal privacy. Don't get me wrong, I am all for using the novel tech for healthcare purposes, but who is to say that this can't be used by nefarious players for mind control or "thought policing" by the State.

If you've been following my articles up to now, you'll begin to perceive, what's apparent to many advanced practitioners of Deep Learning (DL), is the emergence of Game Theoretic concepts in the design of newer architectures. This makes intuitive sense for two reasons. The first intuition is that DL systems will eventually need to tackle situations with imperfect knowledge. In fact, we've already seen this in DeepMind's AlphaGo that uses partial knowledge to tactically and strategically best the world-best human in the game of Go. The second intuition is that systems will not remain monolithic as they are now, but rather would involve multiple coordinating (or competing) cliques of DL systems.

Making decisions in complex environments is a key challenge in artificial intelligence (AI). Situations involving multiple decision makers are particularly complex, leading to computational intractability of principled solution methods. A body of work in AI has tried to mitigate this problem by trying to distill interaction to its essence: how does the policy of one agent influence another agent? If we can find more compact representations of such influence, this can help us deal with the complexity, for instance by searching the space of influences rather than the space of policies. However, so far these notions of influence have been restricted in their applicability to special cases of interaction. In this paper we formalize influence-based abstraction (IBA), which facilitates the elimination of latent state factors without any loss in value, for a very general class of problems described as factored partially observable stochastic games (fPOSGs). On the one hand, this generalizes existing descriptions of influence, and thus can serve as the foundation for improvements in scalability and other insights in decision making in complex multiagent settings. On the other hand, since the presence of other agents can be seen as a generalization of single agent settings, our formulation of IBA also provides a sufficient statistic for decision making under abstraction for a single agent. We also give a detailed discussion of the relations to such previous works, identifying new insights and interpretations of these approaches. In these ways, this paper deepens our understanding of abstraction in a wide range of sequential decision making settings, providing the basis for new approaches and algorithms for a large class of problems.