Demis Hassabis, co-founder and CEO of DeepMind, said: Games and simulations have been a core part of DeepMinds research programme from the very beginning and this approach has already led to significant breakthroughs in AI research. As a former video game designer myself, I couldnt be more excited to be collaborating with Unity, creating virtual environments for developing and testing the kind of smart, flexible algorithms we need to tackle real-world problems. Unity is no stranger to forging thought-leadership in the AI field. In combination with the ML-Agents toolkit, Unity is quickly becoming the platform of choice for the development of intelligent agents. The Unity engine can create the massive simulations needed by researchers to study autonomous artificial agents and develop new kinds of algorithms, which will influence fundamental AI research in areas including robotics, autonomous vehicle development, and many other areas of science and technology.

DeepMind, part of Google parent company Alphabet, is going big on virtual world AI training through a deal with game-making software provider Unity Technologies (which powers games like Monument Valley and Pokémon Go). DeepMind will run the software at a giant scale to train algorithms in physics-realistic environments–part of a growing trend in AI. Game engines like Unity or Unreal provide customizable settings for advanced AI techniques such as reinforcement learning (a kind of machine learning), in which an algorithm pursues a goal through trial and error until it's been mastered. "Games are in many, many ways . . . "You get the visual, the physics, the cognitive, and . . . the social aspect–the interaction."

Training robots to do simple tasks with so-called deep learning has met with limited success, but a San Francisco startup offers a glimmer of hope for future work. Kindred, a three-year-old startup, Thursday offered up a research paper presented at the 2nd Conference on Robot Learning in Zürich, Switzerland. The thrust of the paper is that roboticists need to establish some basic benchmarks about how machine learning, and particularly deep learning, perform before real-world progress can be made. The paper doesn't prove machine learning can teach a robot to move; rather, it suggests there are ways to systematically identify the challenges to doing so, as a basis for future work. In the report, "Benchmarking Reinforcement Learning Algorithms on Real-World Robots," posted on arXiv on September 20th, the authors, A. Rupam Mahmood, Dmytro Korenkevych, Gautham Vasan, William Ma, and James Bergstra, took three commercially available robots and had them move in space to a target location.

Alphabet Inc.'s artificial intelligence group DeepMind Technologies Ltd. said today it will research AI agents with 3-D game development company Unity Technologies Inc., which made the engine for the popular Pokemon Go game. The two companies plan to create a virtual test ground for AI agents that may eventually be used in fields such as autonomous driving and robotics. "DeepMind researchers are trying to crack huge AI problems and Unity provides them with a solution of creating complex virtual environments that will enable the development of algorithms capable of learning to solve complex tasks across diverse environments," Danny Lange, vice president of machine learning and AI at Unity Technologies, said in a statement. "We believe the future of AI is being forged by increasingly sophisticated human-machine interactions, and Unity is proud to be the engine that is enabling these interactions." DeepMind, thanks to its backing from Google-parent Alphabet Inc., has established itself as one of the leading organizations working in AI field, having published more than 200 peer-reviewed papers on the subject in journals such as Nature and Science.

The machine I built costs $3k and has the parts shown below. There's one 1080 Ti GPU to start (you can just as easily use the new 2080 Ti for Machine Learning at $500 more -- just be careful to get one with a blower fan design), a 12 Core CPU, 64GB RAM, and 1TB M.2 SSD. You can add three more GPUs easily for a total of four. Assuming your 1 GPU machine depreciates to $0 in 3 years (very conservative), the chart below shows that if you use it for up to 1 year, it'll be 10x cheaper, including costs for electricity. Amazon discounts pricing if you have a multi-year contract, so the advantage is 4–6x for multi-year contracts.

AI can get pretty good at creating content like images and videos, so researchers are trying to get them to design game levels. Machines are okay working on their own and can regurgitate the same material seen in the numerous training examples fed by its human creators. It's fine if what you're after is more of the same thing, but that's boring for games. Game designing bots need more creativity and the best place to learn is off humans. A team of researchers from the Georgia Institute of Technology conducted a series of experiments where humans partnered up with bots to come up with new levels in Super Mario, a popular Nintendo platform game.

This paper proposes an exploration method for deep reinforcement learning based on parameter space noise. Recent studies have experimentally shown that parameter space noise results in better exploration than the commonly used action space noise. Previous methods devised a way to update the diagonal covariance matrix of a noise distribution and did not consider the direction of the noise vector and its correlation. In addition, fast updates of the noise distribution are required to facilitate policy learning. We propose a method that deforms the noise distribution according to the accumulated returns and the noises that have led to the returns. Moreover, this method switches isotropic exploration and directional exploration in parameter space with regard to obtained rewards. We validate our exploration strategy in the OpenAI Gym continuous environments and modified environments with sparse rewards. The proposed method achieves results that are competitive with a previous method at baseline tasks. Moreover, our approach exhibits better performance in sparse reward environments by exploration with the switching strategy.

The clinical management of several cardiovascular conditions, such as pulmonary hypertension, require the assessment of the right ventricular (RV) function. This work addresses the fully automatic and robust access to one of the key RV biomarkers, its ejection fraction, from the gold standard imaging modality, MRI. The problem becomes the accurate segmentation of the RV blood pool from cine MRI sequences. This work proposes a solution based on Fully Convolutional Neural Networks (FCNN), where our first contribution is the optimal combination of three concepts (the convolution Gated Recurrent Units (GRU), the Generative Adversarial Networks (GAN), and the L1 loss function) that achieves an improvement of 0.05 and 3.49 mm in Dice Index and Hausdorff Distance respectively with respect to the baseline FCNN. This improvement is then doubled by our second contribution, the ROI-GAN, that sets two GANs to cooperate working at two fields of view of the image, its full resolution and the region of interest (ROI). Our rationale here is to better guide the FCNN learning by combining global (full resolution) and local Region Of Interest (ROI) features. The study is conducted in a large in-house dataset of $\sim$ 23.000 segmented MRI slices, and its generality is verified in a publicly available dataset.

Advanced video classification systems decode video frames to derive the necessary texture and motion representations for ingestion and analysis by spatio-temporal deep convolutional neural networks (CNNs). However, when considering visual Internet-of-Things applications, surveillance systems and semantic crawlers of large video repositories, the video capture and the CNN-based semantic analysis parts do not tend to be co-located. This necessitates the transport of compressed video over networks and incurs significant overhead in bandwidth and energy consumption, thereby significantly undermining the deployment potential of such systems. In this paper, we investigate the trade-off between the encoding bitrate and the achievable accuracy of CNN-based video classification models that directly ingest AVC/H.264 and HEVC encoded videos. Instead of retaining entire compressed video bitstreams and applying complex optical flow calculations prior to CNN processing, we only retain motion vector and select texture information at significantly-reduced bitrates and apply no additional processing prior to CNN ingestion. Based on three CNN architectures and two action recognition datasets, we achieve 11%-94% saving in bitrate with marginal effect on classification accuracy. A model-based selection between multiple CNNs increases these savings further, to the point where, if up to 7% loss of accuracy can be tolerated, video classification can take place with as little as 3 kbps for the transport of the required compressed video information to the system implementing the CNN models.

Complex textual information extraction tasks are often posed as sequence labeling or \emph{shallow parsing}, where fields are extracted using local labels made consistent through probabilistic inference in a graphical model with constrained transitions. Recently, it has become common to locally parametrize these models using rich features extracted by recurrent neural networks (such as LSTM), while enforcing consistent outputs through a simple linear-chain model, representing Markovian dependencies between successive labels. However, the simple graphical model structure belies the often complex non-local constraints between output labels. For example, many fields, such as a first name, can only occur a fixed number of times, or in the presence of other fields. While RNNs have provided increasingly powerful context-aware local features for sequence tagging, they have yet to be integrated with a global graphical model of similar expressivity in the output distribution. Our model goes beyond the linear chain CRF to incorporate multiple hidden states per output label, but parametrizes their transitions parsimoniously with low-rank log-potential scoring matrices, effectively learning an embedding space for hidden states. This augmented latent space of inference variables complements the rich feature representation of the RNN, and allows exact global inference obeying complex, learned non-local output constraints. We experiment with several datasets and show that the model outperforms baseline CRF+RNN models when global output constraints are necessary at inference-time, and explore the interpretable latent structure.