In recent years, Google's phones have gone from mediocre cameras to absolute and undeniable kings of mobile photography. This is pretty impressive, but Google has also managed this without adding multiple camera sensors, like most other manufacturers. The Pixel 4 will have two rear-facing camera sensors, which could result in incredible zoom shots if we believe a new teaser.
The key to Google's camera success is machine learning – the same techniques that help an AI beat StarCraft II players or drive a car can feed advanced image processing algorithms. It's no longer the megapixel number of your camera sensors, but the way the software handles all those megapixels after taking a picture.
Claude Zellweger, director of design at Google, recently released a photo taken on Instagram on a Pixel phone (see below). In the comments, he notes that the photo was taken with a 20x zoom. This is better than what you can get with current Pixel phones, even with the maximum digital zoom setting. Naturally, this suggests that Zellweger uses Pixel 4.
Google has done incredible things with its camera sensor isolated on Pixel phones, but you can only do a lot. We know that the Pixels 4 and 4 XL will have a secondary camera on the back and a recent claimed leak (credibly) that will be a 16 MP telephoto camera. The leak did not include the actual zoom level of the lens – we often see 2x and 3x on the phones. However, some recent devices have "periscope" type telephoto devices that can perform 10x optical zoom.
So, how to get to the 20x zoom? It is very very unlikely that this phone has a 20x optical zoom. The photo is crisp and silent, but it's probably the product of Google's excellent image processing. On current Pixel phones, you can use the digital zoom. The phone uses multiple exposures and your natural handshake to refine and sharpen the final image. What we probably see in this image is a 20x digital zoom using a telephoto lens – either something in the 2-3x range or eventually something higher like 10x.
No matter what camera material the Pixel 4 brings to the table, it will likely continue to deliver the best results. We expect the Pixel 4 and 4 XL to land in October at Google's annual hardware event.
In recent years, Intel has been talking about its Cascade Lake servers with DL Boost (also called VNNI, Vector Neural Net Instructions). These new features are a subset of the AVX-512 and are intended to specifically accelerate processor performance in artificial intelligence applications. Historically, many AI applications have favored GPUs over processors. The GPU architecture is much better suited to graphics processors than processors. Processors offer many more thread-based execution resources, but even today's multicore processors are overwhelmed by the parallelism available in a high-end graphics processor core.
Anandtech did you compare the performance of Cascade Lake, the Epyc 7601 (soon outperformed by the AMD Rome 7nm processors, but still today AMD's main server core), and a RTX Titan. The article, written by the excellent Johan De Gelas, discusses different types of neural networks beyond CNN networks (convolutional neural networks), which are generally compared, and explains how a key element of the strategy Intel is competing with Nvidia in workloads where GPUs are not as powerful. or can not yet meet the emerging market needs due to memory capacity constraints (GPUs still can not match CPUs here), the use of 'light' artificial intelligence models does not not requiring long periods of workout, or artificial intelligence models that rely on statistical models of non-neural networks.
Growing revenue from data centers is a critical part of Intel's global strategy for artificial intelligence and machine learning. Nvidia, meanwhile, is keen to protect a market in which it currently has virtually no competition. Intel's Artificial Intelligence strategy is broad and encompasses many products, from Movidius and Nervana to DL Boost on Xeon, to the next GPU Xe range. Nvidia seeks to show that GPUs can be used to handle artificial intelligence calculations in a wider range of workloads. Intel incorporates new artificial intelligence features into its existing products, uses new hardware that, it is hoped, will impact the market, and attempts to create its first serious GPU to challenge the work done by AMD and Nvidia in the consumer market.
Anandtech's benchmarks show, overall, that the gap between Intel and Nvidia remains wide, even with DL Boost. This graph of a recurrent neural network test used a "Long Short Term Memory (LSTM)" network as a neural network. A type of RNN, LSTM "selectively remembers" patterns over a period of time. "Anandtech also used three different configurations to test it – Tensorflow ready to use with conda, a Tensorflow optimized for Intel with PyPi, and optimized version of Tensorflow from the source using Bazel, using the latest version of Tensorflow.
This pair of images captures the relative scale between the processors as well as the comparison with the RTX Titan. Ready-to-use performance was quite poor under AMD, although it improved with the optimized code. Intel's performance skyrocketed like a rocket when the source-optimized version was tested, but even the source-optimized version did not fit the performance of Titan RTX very well. De Gelas notes: "Secondly, we were pretty surprised that our Titan RTX is less than three times faster than our dual Xeon setup," which explains how these comparisons are done in the larger article.
DL Boost is not enough to narrow the gap between Intel and Nvidia, but in all fairness, this has probably never been supposed to be. Intel's goal is to improve AI performance enough on Xeon to make the execution of these plausible workloads on servers that will be mainly used for other purposes, or when creating models of artificial intelligence that do not meet the constraints of D & C A modern graphics processor. The long-term goal of the company is to compete in the AI market with a range of equipment, not just Xeons. Since Xe is not quite ready yet, competing in the HPC space means competing with Xeon for the moment.
For those of you who are wondering about AMD, AMD does not really talk about performing artificial intelligence workloads on Epyc processors, but is focused on its RocM initiative to run CUDA code under OpenCL. AMD does not talk much about this side of its business, but Nvidia dominates the GPU application market for AI and HPC. AMD and Intel both want a piece of space. At the moment, both seem to be fighting for one.
Intel has announced a significant advance for its neuromorphic research processor, named Codehi. The company has now extended its implementation of Loihi to 64 processors, allowing it to create a system with over 8 million neurons (8.3 million). This new configuration (codenamed Pohoiki Beach) delivers 1000 times better performance than conventional processors in applications such as sparse coding, graph search, and constraint satisfaction problems. Intel says the new Pohoiki range offers 10,000 times more energy efficiency than conventional processor architectures in this type of test.
Neuromorphic Computing is a subset of the computer that attempts to imitate the architecture of the brain with the help of modern technological analogues. Instead of implementing a typical CPU clock, for example. Loihi is based on a peaked neural network architecture. The basic Loihi processor contains 128 neuromorphic cores, three Lakefield processor cores (Intel Quark) and an off-chip communications network. Theoretically, Loihi can handle up to 4,096 on-chip cores and 16,384 chips, although Intel has announced that it does not intend to market a design of this size.
"Thanks to the Loihi chip, we were able to demonstrate a power consumption 109 times lower than that of a real-time graphics processor, and a power consumption 5 times lower than that of an inference material. IoT specialized, "said Chris Eliasmith. CEO of Applied Brain Research and Professor at the University of Waterloo. "Even better, while the network is multiplied by 50, Loihi retains performance results in real time and uses 30% more power, while the IoT hardware consumes 500% more power and is no longer in time. real."
The implementation of Pohoiki Beach is not the largest deployment planned for the neuromorphic chip. Intel announces its intention to roll out an even larger concept, dubbed Pohoiki Springs, which will offer "unprecedented levels of performance and efficiency for enhanced neuromorphic workloads."
We covered the advances and research in neuromorphic computing for several years at ET. The work done on these processors is closely related to the work done in AI and the machine intelligence in general, but it's not just about how to perform AI / ML workloads. on existing chips. The ultimate goal is to build processors closer to the human brain.
One of the quirks of computing is that the analogies between the functioning of the human brain and the operation of computers are widespread. Human brains and conventional computers do not overlap very little on their functioning. Transistors are not equivalent to neurons and the pulsed neural network that Loihi uses to transmit information about his own processor cores is intended to be closer to biological processes that humans and other animals use than traditional silicon.
Projects like this one have several long-term research goals, but one of the most fundamental is to better understand how the brain works to replicate some of their energy efficiency. The human brain operates at around 20W. The Exascale supercomputer, considered the minimum for advanced neural simulation of anything more complex than an earthworm, should consume megawatts of power per supercomputer. The difference between these numbers explains why we are primarily interested in the long-term energy efficiency and computing potential of the brain. Architectures such as Loihi are not just an effort to write programs that mimic what humans can do. the goal is to copy aspects of our neurology as well. This makes their progress a little more interesting.
Background image: Tim Herman / Intel Corporation
Steam deploys a new algorithm using the AI to make recommendations on game content. Will it work better than the old beacon system?
The post office Valve presents an automatic learning algorithm to recommend new Steam games appeared first on ExtremeTech.
As artificial intelligence continues to progress rapidly, we still have a long way to go to develop the sensors needed to translate the physical world into computer-based data. While vision and sound were a long way ahead, our other senses have little practical application in the digital world, but this is not the case with robots.
MIT recently created a new robot using GelSight Sensors which allows him to see the objects he touches and to create a 3D map of the texture to better understand it. The video below shows how GelSight technology, generally used for aerospace applications, can "see" what it touches.
GelSight certainly offers an impressive and detailed way of translating the real world into digital information. But that does not make a smart robot, but just very informative fingers that require intelligence to be controlled. Aware of this potential, MIT has created a robot with an artificial intelligence model that exercises on objects that it touches with the help of detailed three-dimensional maps generated by its GelSight sensors. While the robot does not really see what it touches in a traditional optical sense, it receives so much data through its sensors that it can translate that data into visual information and learn from it just like anything else. Ordinary image-oriented convoluted neuron network (CNN).
The MIT robot was trained in 12,000 video recordings of touch data from GelSight sensors, broken down into still image frames, of 200 household objects. Combined with tactile data, this allows the robot to understand the materials that its sensors touch. In one conversation with Engadget, CSAIL Ph.D. student and lead author of this project, explained what their system can now achieve:
"Looking at the scene, our model can imagine the sensation of touching a flat surface or a sharp edge, and by blindly touching, our model can predict interaction with the environment only from tactile feelings. more power to the robot and reduce the data we might need for tasks involving manipulation and seizure of objects. "
The MIT system, which is still in its infancy, is working and it is thanks to their approach. Many researchers and developers in artificial intelligence tend to create models based on the functioning of the human brain, but this often makes no sense. In some cases, we make I want it to work like a human being because his goal is either to approach us or to help us learn more about ourselves by simulating human processes. In most other cases, however, addressing the development of artificial intelligence by imposing a human framework negates the many non-human benefits that software and hardware have to offer.
MIT has chosen to use a much more accurate and capable sensor that any human can only approximate and make the most of the computing power available to AI. By making choices that exploit the benefits of computers rather than forcing humans, they have created a robot that can surpass humans in blind identification tasks. In specific cases, he succeeds already.
Although this does not seem to be the most important problem to solve, touch actually plays an important role in robotics. Niche applications might benefit from a robotic ability to feel the difference between cotton and nylon, but wider applications have much more to offer. For a robot without touching, all objects have the same feeling. He may be able to understand some things visually, but it's rarely helpful.
Think about how you would look after your day if everything you touched felt the same, or more exactly, looked like nothing at all. You would not know what force to use when you plug a cable. You would not be able to understand the practical differences between a printed image of sandpaper and the sandpaper itself.
By providing the robot with a sense of touch and the ability to learn from it, this robot can better judge the materials it touches. He can learn faster and more accurately than he can do simply with standard visuals. He can then use this information to adjust his actions based on the materials that he handles – or at least, it's the ideal goal for the future. If robots can understand touch, they are less likely to cause unintentional damage. At the present time, if you asked most intelligent robots to carry a water balloon, they would not know how to hold it without destroying it. The sense of touch gives robots the ability, through a well trained AI model, to know how to handle different types of objects and to act accordingly.
Although MIT has only created a smarter and smarter robot component, it's still a step in the right direction. A robot designed to understand and incorporate the data it acquires in a tactile way has far better implications for general safety than others. That's how creating backups against potential accidents.
It rarely happens months without learning artificial intelligence dominating the man in a complex game. So it's no surprise that Google's DeepMind masters the winning strategies for Quake III Arena. But unlike past wins in AI, Google's latest approach to enhanced learning has allowed DeepMind to succeed with virtually no instruction and even without its major technical benefits.
Even if you did not already know how to play Capture the flag-The main game mechanism in Quake III Arena: you can understand the rules in less than a minute. Strategic talent, on the other hand, may take some time to develop. If you wanted to program a machine to play even a simple game, it would require a lot more instructions as well as time. Recent developments in artificial intelligence have changed the deal since we can specify the parameters of the artificial neurons as well as the feedback they provide to the machine when performing a task . The machine only knows what actions it can take, whether it fails or not, and this should aim at the goal of failing as seldom as possible. In this particular case, DeepMind can only learn pixels on the screen in the context of these basic settings.
Reinforced learning methods allow the AI to fail often, memorize mistakes and find patterns leading to success. It is quite easy for an AI to succeed without a lot of obstacles and variables, but in a game that requires the cooperation of the team (like Quake III Arena), it must take into account the behavior of the enemy as well as his allies. Winning strategies in team games rarely involve a single player. The beginnings of Michael Jordan's basketball career clearly show how to play the role of a player who plays for himself will not lead a team to victory. But the AI is not cluttered with conflicting goals. In about 450,000 games, about four years of practice for a human, DeepMind intuitively intuited successful team-based strategies that did not win, but that made it possible to win against skilled human players far more often than 'he had lost.
Google used this training data to create DeepMind's "For the Win" (FTW) agents to play as individual team members in Quake III Arena. In each game played, Google randomly assigned teams of an equal number of human players and FTW agents. FTW agents have managed a likely "winning rate" of about 1.23 times that of the most powerful players. Playing with average human players, this victory rate has risen to around 1.5x. Of course, machines have a decisive advantage in terms of speed of processing and detailed information in memory. Nevertheless, even the introduction of a normal delay of 257 milliseconds only caused the loss of FTW agents against competent players around 79% of the time.
DeepMind FTW agents owe their success to some essential elements of the enhanced learning process. As long as no instructions were provided, no neurons were coded to respond to specific game events, such as the capture of an agent's flag or when a teammate had a flag to calculate the context of these events. Because all learning is done visually, the arrangement of artificial neurons has been modeled on the visual cortex of the human brain. Two long term memory Networks (LTSMs), each operating on different time scales, process visual data with their own learning objectives. This dual concurrent process offers each FTW agent the advantage of comparing the opportunities taken at the machine equivalent of different perspectives. Agents determine their choices based on the outcome of this process and play the game by emulating a game controller. As you can see in the video above, the quick moves offer a distinct advantage and illustrate a distinct style of play that few, if any, humans could handle.
In face-to-face matches, the superiority of AI may seem to be an insurmountable obstacle even for the best players. In a team environment, however, AI and humans can actually work together and The competition so as not to sacrifice the pleasure of the game.
VentureBeat talked to Thore Graepel, DeepMind Scientist and Professor of Computer Science at Global University London, who explains in more detail the benefits of these efforts:
Our results demonstrate that multi-agent reinforcement learning can successfully tackle a complex game to the point that human players even think that computer gamers are better teammates. They also provide a fascinating in-depth analysis of the behavior of trained agents, their collaboration and the representation of their environment. What makes these results so interesting is that these agents perceive their environment in the first person, as would a human player. To learn to play tactically and to collaborate with their teammates, these agents must rely on the information provided by the game results – without the teacher or coach telling them what to do.
These efforts provide a more optimistic look at how humans and artificial intelligence can coexist in a beneficial way. Although it can not relieve some of the most important concerns raised by Amnesty International about the near future, these positive examples help determine the best ways to use this powerful new technology.