Content Threat Removal (CTR) is a cybersecurity technology intended to defeat the threat posed by handling digital content in the cyberspace. Unlike other defenses, including antivirus software and sandboxed execution, CTR does not rely on being able to detect threats. Similar to Content Disarm and Reconstruction, CTR is designed to remove the threat without knowing whether it has done so and acts without knowing if data contains a threat or not. Detection strategies work by detecting unsafe content, and then blocking or removing that content. Content that is deemed safe is delivered to its destination. In contrast, Content Threat Removal assumes all data is hostile and delivers none of it to the destination, regardless of whether it is actually hostile. Although no data is delivered, the business information carried by the data is delivered using new data created for the purpose. == Threat == Advanced attacks continuously defeat defenses that are based on detection. These are often referred to as zero-day attacks, because as soon as they are discovered attack detection mechanisms must be updated to identify and neutralize the attack, and until they are, all systems are unprotected. These attacks succeed because attackers find new ways of evading detection. Polymorphic code can be used to evade the detection of known unsafe data and sandbox detection allows attacks to evade dynamic analysis. == Method == A Content Threat Removal defence works by intercepting data on its way to its destination. The business information carried by the data is extracted and the data is discarded. Then entirely new, clean and safe data is built to carry the information to its destination. The effect of building new data to carry the business information is that any unsafe elements of the original data are left behind and discarded. This includes executable data, macros, scripts and malformed data that trigger vulnerabilities in applications. While CTR is a form of content transformation, not all transformations provide a complete defence against the content threat. == Applicability == CTR is applicable to user-to-user traffic, such as email and chat, and machine-to-machine traffic, such as web services. Data transfers can be intercepted by in-line application layer proxies and these can transform the way information content is delivered to remove any threat. CTR works by extracting business information from data and it is not possible to extract information from executable code. This means CTR is not directly applicable to web browsing, since most web pages are code. It can, however, be applied to content that is downloaded from, and uploaded to, websites. Although most web pages cannot be transformed to render them safe, web browsing can be isolated and the remote access protocols used to reach the isolated environment can be subjected to CTR. CTR provides a solution to the problem of stegware. It naturally removes detectable steganography and eliminates symbiotic and permutation steganography through normalisation.
Real-time computer graphics
Real-time computer graphics or real-time rendering is the sub-field of computer graphics focused on producing and analyzing images in real time. The term can refer to anything from rendering an application's graphical user interface (GUI) to real-time image analysis, but is most often used in reference to interactive 3D computer graphics, typically using a graphics processing unit (GPU). One example of this concept is a video game that rapidly renders changing 3D environments to produce an illusion of motion. Computers have been capable of generating 2D images such as simple lines, images and polygons in real time since their invention. However, quickly rendering detailed 3D objects is a daunting task for traditional Von Neumann architecture-based systems. An early workaround to this problem was the use of sprites, 2D images that could imitate 3D graphics. Different techniques for rendering now exist, such as ray-tracing and rasterization. Using these techniques and advanced hardware, computers can now render images quickly enough to create the illusion of motion while simultaneously accepting user input. This means that the user can respond to rendered images in real time, producing an interactive experience. == Principles of real-time 3D computer graphics == The goal of computer graphics is to generate computer-generated images, or frames, using certain desired metrics. One such metric is the number of frames generated in a given second. Real-time computer graphics systems differ from traditional (i.e., non-real-time) rendering systems in that non-real-time graphics typically rely on ray tracing. In this process, millions or billions of rays are traced from the camera to the world for detailed rendering—this expensive operation can take hours or days to render a single frame. Real-time graphics systems must render each image in less than 1/30th of a second. Ray tracing is far too slow for these systems; instead, they employ the technique of z-buffer triangle rasterization. In this technique, every object is decomposed into individual primitives, usually triangles. Each triangle gets positioned, rotated and scaled on the screen, and rasterizer hardware (or a software emulator) generates pixels inside each triangle. These triangles are then decomposed into atomic units called fragments that are suitable for displaying on a display screen. The fragments are drawn on the screen using a color that is computed in several steps. For example, a texture can be used to "paint" a triangle based on a stored image, and then shadow mapping can alter that triangle's colors based on line-of-sight to light sources. === Video game graphics === Real-time graphics optimizes image quality subject to time and hardware constraints. GPUs and other advances increased the image quality that real-time graphics can produce. GPUs are capable of handling millions of triangles per frame, and modern DirectX/OpenGL class hardware is capable of generating complex effects, such as shadow volumes, motion blurring, and triangle generation, in real-time. The advancement of real-time graphics is evidenced in the progressive improvements between actual gameplay graphics and the pre-rendered cutscenes traditionally found in video games. Cutscenes are typically rendered in real-time—and may be interactive. Although the gap in quality between real-time graphics and traditional off-line graphics is narrowing, offline rendering remains much more accurate. === Advantages === Real-time graphics are typically employed when interactivity (e.g., player feedback) is crucial. When real-time graphics are used in films, the director has complete control of what has to be drawn on each frame, which can sometimes involve lengthy decision-making. Teams of people are typically involved in the making of these decisions. In real-time computer graphics, the user typically operates an input device to influence what is about to be drawn on the display. For example, when the user wants to move a character on the screen, the system updates the character's position before drawing the next frame. Usually, the display's response-time is far slower than the input device—this is justified by the immense difference between the (fast) response time of a human being's motion and the (slow) perspective speed of the human visual system. This difference has other effects too: because input devices must be very fast to keep up with human motion response, advancements in input devices (e.g., the current Wii remote) typically take much longer to achieve than comparable advancements in display devices. Another important factor controlling real-time computer graphics is the combination of physics and animation. These techniques largely dictate what is to be drawn on the screen—especially where to draw objects in the scene. These techniques help realistically imitate real world behavior (the temporal dimension, not the spatial dimensions), adding to the computer graphics' degree of realism. Real-time previewing with graphics software, especially when adjusting lighting effects, can increase work speed. Some parameter adjustments in fractal generating software may be made while viewing changes to the image in real time. == Rendering pipeline == The graphics rendering pipeline ("rendering pipeline" or simply "pipeline") is the foundation of real-time graphics. Its main function is to render a two-dimensional image in relation to a virtual camera, three-dimensional objects (an object that has width, length, and depth), light sources, lighting models, textures and more. === Architecture === The architecture of the real-time rendering pipeline can be divided into conceptual stages: application, geometry and rasterization. === Application stage === The application stage is responsible for generating "scenes", or 3D settings that are drawn to a 2D display. This stage is implemented in software that developers optimize for performance. This stage may perform processing such as collision detection, speed-up techniques, animation and force feedback, in addition to handling user input. Collision detection is an example of an operation that would be performed in the application stage. Collision detection uses algorithms to detect and respond to collisions between (virtual) objects. For example, the application may calculate new positions for the colliding objects and provide feedback via a force feedback device such as a vibrating game controller. The application stage also prepares graphics data for the next stage. This includes texture animation, animation of 3D models, animation via transforms, and geometry morphing. Finally, it produces primitives (points, lines, and triangles) based on scene information and feeds those primitives into the geometry stage of the pipeline. === Geometry stage === The geometry stage manipulates polygons and vertices to compute what to draw, how to draw it and where to draw it. Usually, these operations are performed by specialized hardware or GPUs. Variations across graphics hardware mean that the "geometry stage" may actually be implemented as several consecutive stages. ==== Model and view transformation ==== Before the final model is shown on the output device, the model is transformed onto multiple spaces or coordinate systems. Transformations move and manipulate objects by altering their vertices. Transformation is the general term for the four specific ways that manipulate the shape or position of a point, line or shape. ==== Lighting ==== In order to give the model a more realistic appearance, one or more light sources are usually established during transformation. However, this stage cannot be reached without first transforming the 3D scene into view space. In view space, the observer (camera) is typically placed at the origin. If using a right-handed coordinate system (which is considered standard), the observer looks in the direction of the negative z-axis with the y-axis pointing upwards and the x-axis pointing to the right. ==== Projection ==== Projection is a transformation used to represent a 3D model in a 2D space. The two main types of projection are orthographic projection (also called parallel) and perspective projection. The main characteristic of an orthographic projection is that parallel lines remain parallel after the transformation. Perspective projection utilizes the concept that if the distance between the observer and model increases, the model appears smaller than before. Essentially, perspective projection mimics human sight. ==== Clipping ==== Clipping is the process of removing primitives that are outside of the view box in order to facilitate the rasterizer stage. Once those primitives are removed, the primitives that remain will be drawn into new triangles that reach the next stage. ==== Screen mapping ==== The purpose of screen mapping is to find out the coordinates of the primitives during the clipping stage. ==== Rasterizer stage ==== The rasterizer
VoID
The Vocabulary of Interlinked Datasets (VoID) is a vocabulary for providing concise summaries (metadata) of Resource Description Framework (RDF) datasets—meaningful collections of semantic triples—using the syntax of RDF Schema. It can be used for general metadata (such as information about the license of the dataset), access metadata (information about how to access the dataset), structural metadata (information about how the dataset is structured), and linking metadata (information about links between datasets). A linked dataset is a collection of data, published and maintained by a single provider, available as RDF on the Web, where at least some of the resources in the dataset are identified by dereferencable Uniform Resource Identifiers (URIs). VoID is used to provide metadata on RDF datasets to facilitate query processing on a graph of interlinked datasets in the Semantic Web.
IEEE Transactions on Pattern Analysis and Machine Intelligence
IEEE Transactions on Pattern Analysis and Machine Intelligence (sometimes abbreviated as IEEE PAMI or simply PAMI) is a monthly peer-reviewed scientific journal published by the IEEE Computer Society. == Background == The journal covers research in computer vision and image understanding, pattern analysis and recognition, machine intelligence, machine learning, search techniques, document and handwriting analysis, medical image analysis, video and image sequence analysis, content-based retrieval of image and video, and face and gesture recognition. The editor-in-chief is Kyoung Mu Lee (Seoul National University). According to the Journal Citation Reports, the journal has a 2023 impact factor of 20.8.
Pattern language
A pattern language is an organized and coherent set of patterns, each of which describes a problem and the core of a solution that can be used in many ways within a specific field of expertise. The term was coined by architect Christopher Alexander and popularized by his 1977 book A Pattern Language. A pattern language can also be an attempt to express the deeper wisdom of what brings aliveness within a particular field of human endeavor, through a set of interconnected patterns. Aliveness is one placeholder term for "the quality that has no name": a sense of wholeness, spirit, or grace, that while of varying form, is precise and empirically verifiable. Alexander claims that ordinary people can use this design approach to successfully solve very large, complex design problems. == What is a pattern? == When a designer designs something – whether a house, computer program, or lamp – they must make many decisions about how to solve problems. A single problem is documented with its typical place (the syntax), and use (the grammar) with the most common and recognized good solution seen in the wild, like the examples seen in dictionaries. Each such entry is a single design pattern. Each pattern has a name, a descriptive entry, and some cross-references, much like a dictionary entry. A documented pattern should explain why that solution is good in the pattern's contexts. Elemental or universal patterns such as "door" or "partnership" are versatile ideals of design, either as found in experience or for use as components in practice, explicitly described as holistic resolutions of the forces in recurrent contexts and circumstances, whether in architecture, medicine, software development or governance, etc. Patterns might be invented or found and studied, such as the naturally occurring patterns of design that characterize human environments. Like all languages, a pattern language has vocabulary, syntax, and grammar – but a pattern language applies to some complex activity other than communication. In pattern languages for design, the parts break down in this way: The language description – the vocabulary – is a collection of named, described solutions to problems in a field of interest. These are called design patterns. So, for example, the language for architecture describes items like: settlements, buildings, rooms, windows, latches, etc. Each solution includes syntax, a description that shows where the solution fits in a larger, more comprehensive or more abstract design. This automatically links the solution into a web of other needed solutions. For example, rooms have ways to get light, and ways to get people in and out. The solution includes grammar that describes how the solution solves a problem or produces a benefit. So, if the benefit is unneeded, the solution is not used. Perhaps that part of the design can be left empty to save money or other resources; if people do not need to wait to enter a room, a simple doorway can replace a waiting room. In the language description, grammar and syntax cross index (often with a literal alphabetic index of pattern names) to other named solutions, so the designer can quickly think from one solution to related, needed solutions, and document them in a logical way. In Christopher Alexander's book A Pattern Language, the patterns are in decreasing order by size, with a separate alphabetic index. The web of relationships in the index of the language provides many paths through the design process. This simplifies the design work because designers can start the process from any part of the problem they understand and work toward the unknown parts. At the same time, if the pattern language has worked well for many projects, there is reason to believe that even a designer who does not completely understand the design problem at first will complete the design process, and the result will be usable. For example, skiers coming inside must shed snow and store equipment. The messy snow and boot cleaners should stay outside. The equipment needs care, so the racks should be inside. == Many patterns form a language == Just as words must have grammatical and semantic relationships to each other in order to make a spoken language useful, design patterns must be related to each other in position and utility order to form a pattern language. Christopher Alexander's work describes a process of decomposition, in which the designer has a problem (perhaps a commercial assignment), selects a solution, then discovers new, smaller problems resulting from the larger solution. Occasionally, the smaller problems have no solution, and a different larger solution must be selected. Eventually all of the remaining design problems are small enough or routine enough to be solved by improvisation by the builders, and the "design" is done. The actual organizational structure (hierarchical, iterative, etc.) is left to the discretion of the designer, depending on the problem. This explicitly lets a designer explore a design, starting from some small part. When this happens, it's common for a designer to realize that the problem is actually part of a larger solution. At this point, the design almost always becomes a better design. In the language, therefore, each pattern has to indicate its relationships to other patterns and to the language as a whole. This gives the designer using the language a great deal of guidance about the related problems that must be solved. The most difficult part of having an outside expert apply a pattern language is in fact to get a reliable, complete list of the problems to be solved. Of course, the people most familiar with the problems are the people that need a design. So, Alexander famously advocated on-site improvisation by concerned, empowered users, as a powerful way to form very workable large-scale initial solutions, maximizing the utility of a design, and minimizing the design rework. The desire to empower users of architecture was, in fact, what led Alexander to undertake a pattern language project for architecture in the first place. == Design problems in a context == An important aspect of design patterns is to identify and document the key ideas that make a good system different from a poor system (that may be a house, a computer program or an object of daily use), and to assist in the design of future systems. The idea expressed in a pattern should be general enough to be applied in very different systems within its context, but still specific enough to give constructive guidance. The range of situations in which the problems and solutions addressed in a pattern apply is called its context. An important part in each pattern is to describe this context. Examples can further illustrate how the pattern applies to very different situation. For instance, Alexander's pattern "A PLACE TO WAIT" addresses bus stops in the same way as waiting rooms in a surgery, while still proposing helpful and constructive solutions. The "Gang-of-Four" book Design Patterns by Gamma et al. proposes solutions that are independent of the programming language, and the program's application domain. Still, the problems and solutions described in a pattern can vary in their level of abstraction and generality on the one side, and specificity on the other side. In the end this depends on the author's preferences. However, even a very abstract pattern will usually contain examples that are, by nature, absolutely concrete and specific. Patterns can also vary in how far they are proven in the real world. Alexander gives each pattern a rating by zero, one or two stars, indicating how well they are proven in real-world examples. It is generally claimed that all patterns need at least some existing real-world examples. It is, however, conceivable to document yet unimplemented ideas in a pattern-like format. The patterns in Alexander's book also vary in their level of scale – some describing how to build a town or neighbourhood, others dealing with individual buildings and the interior of rooms. Alexander sees the low-scale artifacts as constructive elements of the large-scale world, so they can be connected to a hierarchic network. === Balancing of forces === A pattern must characterize the problems that it is meant to solve, the context or situation where these problems arise, and the conditions under which the proposed solutions can be recommended. Often these problems arise from a conflict of different interests or "forces". A pattern emerges as a dialogue that will then help to balance the forces and finally make a decision. For instance, there could be a pattern suggesting a wireless telephone. The forces would be the need to communicate, and the need to get other things done at the same time (cooking, inspecting the bookshelf). A very specific pattern would be just "WIRELESS TELEPHONE". More general patterns would be "WIRELESS DEVICE" or "SECONDARY ACTIVITY", suggesting that a secondary activity (such as talking on t
Instance selection
Instance selection (or dataset reduction, or dataset condensation) is an important data pre-processing step that can be applied in many machine learning (or data mining) tasks. Approaches for instance selection can be applied for reducing the original dataset to a manageable volume, leading to a reduction of the computational resources that are necessary for performing the learning process. Algorithms of instance selection can also be applied for removing noisy instances, before applying learning algorithms. This step can improve the accuracy in classification problems. Algorithm for instance selection should identify a subset of the total available data to achieve the original purpose of the data mining (or machine learning) application as if the whole data had been used. Considering this, the optimal outcome of IS would be the minimum data subset that can accomplish the same task with no performance loss, in comparison with the performance achieved when the task is performed using the whole available data. Therefore, every instance selection strategy should deal with a trade-off between the reduction rate of the dataset and the classification quality. == Instance selection algorithms == The literature provides several different algorithms for instance selection. They can be distinguished from each other according to several different criteria. Considering this, instance selection algorithms can be grouped in two main classes, according to what instances they select: algorithms that preserve the instances at the boundaries of classes and algorithms that preserve the internal instances of the classes. Within the category of algorithms that select instances at the boundaries it is possible to cite DROP3, ICF and LSBo. On the other hand, within the category of algorithms that select internal instances, it is possible to mention ENN and LSSm. In general, algorithm such as ENN and LSSm are used for removing harmful (noisy) instances from the dataset. They do not reduce the data as the algorithms that select border instances, but they remove instances at the boundaries that have a negative impact on the data mining task. They can be used by other instance selection algorithms, as a filtering step. For example, the ENN algorithm is used by DROP3 as the first step, and the LSSm algorithm is used by LSBo. There is also another group of algorithms that adopt different selection criteria. For example, the algorithms LDIS, CDIS and XLDIS select the densest instances in a given arbitrary neighborhood. The selected instances can include both, border and internal instances. The LDIS and CDIS algorithms are very simple and select subsets that are very representative of the original dataset. Besides that, since they search by the representative instances in each class separately, they are faster (in terms of time complexity and effective running time) than other algorithms, such as DROP3 and ICF. Besides that, there is a third category of algorithms that, instead of selecting actual instances of the dataset, select prototypes (that can be synthetic instances). In this category it is possible to include PSSA, PSDSP and PSSP. The three algorithms adopt the notion of spatial partition (a hyperrectangle) for identifying similar instances and extract prototypes for each set of similar instances. In general, these approaches can also be modified for selecting actual instances of the datasets. The algorithm ISDSP adopts a similar approach for selecting actual instances (instead of prototypes).
Google Brain
Google Brain was a deep learning artificial intelligence research team that served as the sole AI branch of Google before being incorporated under the newer umbrella of Google AI, a research division at Google dedicated to artificial intelligence. Formed in 2011, it combined open-ended machine learning research with information systems and large-scale computing resources. It created tools such as TensorFlow, which allow neural networks to be used by the public, and multiple internal AI research projects, and aimed to create research opportunities in machine learning and natural language processing. It was merged into former Google sister company DeepMind to form Google DeepMind in April 2023. == History == The Google Brain project began in 2011 as a part-time research collaboration between Google fellow Jeff Dean and Google Researcher Greg Corrado. Google Brain started as a Google X project and became so successful that it was graduated back to Google: Astro Teller has said that Google Brain paid for the entire cost of Google X. In June 2012, The New York Times reported that a cluster of 16,000 processors in 1,000 computers dedicated to mimicking some aspects of human brain activity had successfully trained itself to recognize a cat based on 10 million digital images taken from YouTube videos. The story was also covered by National Public Radio (NPR). In March 2013, Google hired Geoffrey Hinton, a leading researcher in the deep learning field, and acquired the company DNNResearch Inc. headed by Hinton. Hinton said that he would be dividing his future time between his university research and his work at Google. In April 2023, Google Brain merged with Google sister company DeepMind to form Google DeepMind, as part of the company's continued efforts to accelerate work on AI. == Team and location == Google Brain was initially established by Google Fellow Jeff Dean and visiting Stanford professor Andrew Ng. In 2014, the team included Jeff Dean, Quoc V. Le, Ilya Sutskever, Alex Krizhevsky, Samy Bengio, and Vincent Vanhoucke. In 2017, team members included Anelia Angelova, Samy Bengio, Greg Corrado, George Dahl, Michael Isard, Anjuli Kannan, Hugo Larochelle, Chris Olah, Benoit Steiner, Vincent Vanhoucke, Vijay Vasudevan, and Fernanda Viegas. Chris Lattner, who created Apple's programming language Swift and then ran Tesla's autonomy team for six months, joined Google Brain's team in August 2017. Lattner left the team in January 2020 and joined SiFive. As of 2021, Google Brain was led by Jeff Dean, Geoffrey Hinton, and Zoubin Ghahramani. Other members include Katherine Heller, Pi-Chuan Chang, Ian Simon, Jean-Philippe Vert, Nevena Lazic, Anelia Angelova, Lukasz Kaiser, Carrie Jun Cai, Eric Breck, Ruoming Pang, Carlos Riquelme, Hugo Larochelle, and David Ha. Samy Bengio left the team in April 2021, and Zoubin Ghahramani took on his responsibilities. Google Research includes Google Brain and is based in Mountain View. It also has satellite groups in Accra, Amsterdam, Atlanta, Beijing, Berlin, Cambridge, Israel, Los Angeles, London, Montreal, Munich, New York City, Paris, Pittsburgh, Princeton, San Francisco, Seattle, Tokyo, Toronto, and Zurich. == Projects == === Artificial-intelligence-devised encryption system === In October 2016, Google Brain designed an experiment to determine that neural networks are capable of learning secure symmetric encryption. In this experiment, three neural networks were created: Alice, Bob and Eve. Adhering to the idea of a generative adversarial network (GAN), the goal of the experiment was for Alice to send an encrypted message to Bob that Bob could decrypt, but the adversary, Eve, could not. Alice and Bob maintained an advantage over Eve, in that they shared a key used for encryption and decryption. In doing so, Google Brain demonstrated the capability of neural networks to learn secure encryption. === Image enhancement === In February 2017, Google Brain determined a probabilistic method for converting pictures with 8x8 resolution to a resolution of 32x32. The method built upon an already existing probabilistic model called pixelCNN to generate pixel translations. The proposed software utilizes two neural networks to make approximations for the pixel makeup of translated images. The first network, known as the "conditioning network," downsizes high-resolution images to 8x8 and attempts to create mappings from the original 8x8 image to these higher-resolution ones. The other network, known as the "prior network," uses the mappings from the previous network to add more detail to the original image. The resulting translated image is not the same image in higher resolution, but rather a 32x32 resolution estimation based on other existing high-resolution images. Google Brain's results indicate the possibility for neural networks to enhance images. === Google Translate === The Google Brain contributed to the Google Translate project by employing a new deep learning system that combines artificial neural networks with vast databases of multilingual texts. In September 2016, Google Neural Machine Translation (GNMT) was launched, an end-to-end learning framework, able to learn from a large number of examples. Previously, Google Translate's Phrase-Based Machine Translation (PBMT) approach would statistically analyze word by word and try to match corresponding words in other languages without considering the surrounding phrases in the sentence. But rather than choosing a replacement for each individual word in the desired language, GNMT evaluates word segments in the context of the rest of the sentence to choose more accurate replacements. Compared to older PBMT models, the GNMT model scored a 24% improvement in similarity to human translation, with a 60% reduction in errors. The GNMT has also shown significant improvement for notoriously difficult translations, like Chinese to English. While the introduction of the GNMT has increased the quality of Google Translate's translations for the pilot languages, it was very difficult to create such improvements for all of its 103 languages. Addressing this problem, the Google Brain Team was able to develop a Multilingual GNMT system, which extended the previous one by enabling translations between multiple languages. Furthermore, it allows for Zero-Shot Translations, which are translations between two languages that the system has never explicitly seen before. Google announced that Google Translate can now also translate without transcribing, using neural networks. This means that it is possible to translate speech in one language directly into text in another language, without first transcribing it to text. According to the Researchers at Google Brain, this intermediate step can be avoided using neural networks. In order for the system to learn this, they exposed it to many hours of Spanish audio together with the corresponding English text. The different layers of neural networks, replicating the human brain, were able to link the corresponding parts and subsequently manipulate the audio waveform until it was transformed to English text. Another drawback of the GNMT model is that it causes the time of translation to increase exponentially with the number of words in the sentence. This caused the Google Brain Team to add 2000 more processors to ensure the new translation process would still be fast and reliable. === Robotics === Aiming to improve traditional robotics control algorithms where new skills of a robot need to be hand-programmed, robotics researchers at Google Brain are developing machine learning techniques to allow robots to learn new skills on their own. They also attempt to develop ways for information sharing between robots so that robots can learn from each other during their learning process, also known as cloud robotics. As a result, Google has launched the Google Cloud Robotics Platform for developers in 2019, an effort to combine robotics, AI, and the cloud to enable efficient robotic automation through cloud-connected collaborative robots. Robotics research at Google Brain has focused mostly on improving and applying deep learning algorithms to enable robots to complete tasks by learning from experience, simulation, human demonstrations, and/or visual representations. For example, Google Brain researchers showed that robots can learn to pick and throw rigid objects into selected boxes by experimenting in an environment without being pre-programmed to do so. In another research, researchers trained robots to learn behaviors such as pouring liquid from a cup; robots learned from videos of human demonstrations recorded from multiple viewpoints. Google Brain researchers have collaborated with other companies and academic institutions on robotics research. In 2016, the Google Brain Team collaborated with researchers at X in a research on learning hand-eye coordination for robotic grasping. Their method allowed real-time robot control for grasping novel objec