We consider reinforcement learning in input-driven environments, where an exogenous, stochastic input process affects the dynamics of the system. Input processes arise in many applications, including queuing systems, robotics control with disturbances, and object tracking. Since the state dynamics and rewards depend on the input process, the state alone provides limited information for the expected future returns. Therefore, policy gradient methods with standard state-dependent baselines suffer high variance during training. We derive a bias-free, input-dependent baseline to reduce this variance, and analytically show its benefits over state-dependent baselines. We then propose a meta-learning approach to overcome the complexity of learning a baseline that depends on a long sequence of inputs. Our experimental results show that across environments from queuing systems, computer networks, and MuJoCo robotic locomotion, input-dependent baselines consistently improve training stability and result in better eventual policies.

Estimating a large alphabet probability distribution from a limited number of samples is a fundamental problem in machine learning and statistics. A variety of estimation schemes have been proposed over the years, mostly inspired by the early work of Laplace and the seminal contribution of Good and Turing. One of the basic assumptions shared by most commonly-used estimators is the unique correspondence between the symbol's sample frequency and its estimated probability. In this work we tackle this paradigmatic assumption; we claim that symbols with "similar" frequencies shall be assigned the same estimated probability value. This way we regulate the number of parameters and improve generalization. In this preliminary report we show that by applying an ensemble of such regulated estimators, we introduce a dramatic enhancement in the estimation accuracy (typically up to 50%), compared to currently known methods. An implementation of our suggested method is publicly available at the first author's web-page.

If AI continues to develop exponentially it will offer significant productivity gains for all sectors of the economy. Crypto offers AI significant advantages versus traditional sources of data for reasons I explain below; and as such, one of the most impactful ways AI will affect our economy will be when it is deployed on top of cryptographic'rails' that define everything from capital raising through to underlying economic activity at the transactional level. Crypto will effectively become the protocol for AI to talk to other AI. Capital will follow where data is abundant and trusted. This will have profound effects on the functioning of capital markets and the theory of the firm over the course of this century.

Baidu marked volume production of Apolong autonomous buses, developed with Chinese busmaker King Long and using Apollo autonomous driving software, for commercial operation in cities including Beijing, Shenzhen, Pingtan and Wuhan, the company said at its developer conference in Beijing on July 4, 2018. Intel is getting a big opportunity to show off advanced vision technology and software from its Mobileye unit by partnering with Baidu's Apollo, the Chinese tech giant's autonomous driving consortium. Working with Mobileye, Baidu will integrate the Israeli company's Responsibility Sensitive Safety software into both its open-source Apollo platform shared with partner companies and commercial Apollo Pilot programs, Intel said in a statement. Baidu will also use Mobileye's camera-based surround computer vision system for automated vehicles being developed for the Chinese market. "Our team recognizes the value and critical role that Mobileye's RSS model plays in safely deploying autonomous driving," Weihao Gu, general manager of Baidu's Intelligent Driving Unit, said in a statement.

Just a quick video to let you know that Japan has just announced that it is going to make everyone, even primary school students, AI-literate within the next few years. The Japanese government wants primary school children to be as familiar with AI as they are with the 3Rs. Facing a severe skills shortage of technicians, Japan is having to rethink its entire tertiary system and government governance of technical infrastructure, which includes a massive reskilling operation. AI is deemed to be a key plank of Japan's economy in the near future. But skilled AI technicians are in short supply and Japan's immigration policies are being looked at with a view to opening the country up to foreign workers.

The artificial intelligence in supply chain market is likely to grow at a CAGR of 45.55% between 2018 and 2025. The artificial intelligence in supply chain market is expected to reach USD 10,110.2 million by 2025 from USD 730.6 million in 2018, at a CAGR of 45.55%. Growth in this market can largely be attributed to factors such as growing big data, demand for greater visibility and transparency into supply chain data and processes, and adoption of AI for improving consumer services and their satisfaction. On the other hand, the limited number of artificial intelligence technology experts is expected to restrict adoption, which in turn may limit market growth to a certain extent. The market for software offerings is expected to hold a largest share during the forecast period.

The shape of a particle of volcanic ash can help volcanologists determine the type of eruption that produced it, which can help response teams know how to react in the aftermath of an eruption. For example, it might let them know how large of an area to evacuate. But in the past, categorizing ash was time-consuming, subjective, and reliant on the availability of highly trained experts. Now, scientists from the Earth-Life Science Institute at the Tokyo Institute of Technology have developed an artificial intelligence (AI) program that analyzes volcanic ash particles to determine their shape. The scientists detail their research in a paper published in May in the journal Scientific Reports.

I expected my summer engineering internship to include things like updating old 3-D models, creating part designs, and learning the ins and outs of how a company works. I didn't expect it to involve learning to make my colleagues obsolete. It was the summer after my sophomore year of college, at a company in Southern California. At the beginning of the internship, my manager asked me to implement 3-D printing to streamline a complicated mold-making process. I have long been obsessed with 3-D printing (I own two machines myself), so I was thrilled to introduce it into the business.

We asked young scientists this question: How do broad interests benefit your science? Scientists with a variety of hobbies responded that their extracurricular activities have enhanced a wide range of skills, from creativity to communication to resilience. Many also mentioned the value of clearing their minds and relaxing. Follow NextGen and share your own hobbies on Twitter with #NextGenSci. As a rock climber, you have to risk falling in order to become better; the same principle applies in science.

Gaming culture is rife with graphic representations of gun violence and has been since arcade goers first blew aliens out of Space Invader's skies. You'll be hard-pressed to find more than a handful of AAA titles designed for adults (sit down Rayman) that don't rely on firearms, or use gore in substitution, either as a primary tool for the gameplay or as a thematic element. While firearms have been a mainstay in video games since the mid '70s, few games have cemented their position in popular culture than 1993's pioneering first person shooter, Doom. Not only did Doom bring the FPS genre into mainstream gaming culture, they also helped to normalize a level of gore not often seen since 1988's Turbografx 16 hit, Splatterhouse, one of the first games to ever carry a parental warning label. Doom's influence is clear in nearly every game in the genre.