NeuReality raises $8M for its novel AI inferencing platform

NeuReality, an Israeli AI hardware startup that is working on a novel approach to improving AI inferencing platforms by doing away with the current CPU-centric model, is coming out of stealth today and announcing an $8 million seed round. The group of investors includes Cardumen Capital, crowdfunding platform OurCrowd and Varana Capital. The company also today announced that Naveen Rao, the GM of Intel’s AI Products Group and former CEO of Nervana System (which Intel acquired), is joining the company’s board of directors.

The founding team, CEO Moshe Tanach, VP of operations Tzvika Shmueli and VP for very large-scale integration Yossi Kasus, has a background in AI but also networking, with Tanach spending time at Marvell and Intel, for example, Shmueli at Mellanox and Habana Labs and Kasus at Mellanox, too.

It’s the team’s networking and storage knowledge and seeing how that industry built its hardware that now informs how NeuReality is thinking about building its own AI platform. In an interview ahead of today’s announcement, Tanach wasn’t quite ready to delve into the details of NeuReality’s architecture, but the general idea here is to build a platform that will allow hyperscale clouds and other data center owners to offload their ML models to a far more performant architecture where the CPU doesn’t become a bottleneck.

“We kind of combined a lot of techniques that we brought from the storage and networking world,” Tanach explained. Think about traffic manager and what it does for Ethernet packets. And we applied it to AI. So we created a bottom-up approach that is built around the engine that you need. Where today, they’re using neural net processors — we have the next evolution of AI computer engines.”

As Tanach noted, the result of this should be a system that — in real-world use cases that include not just synthetic benchmarks of the accelerator but also the rest of the overall architecture — offer 15 times the performance per dollar for basic deep learning offloading and far more once you offload the entire pipeline to its platform.

NeuReality is still in its early days, and while the team has working prototypes now, based on a Xilinx FPGA, it expects to be able to offer its fully custom hardware solution early next year. As its customers, NeuReality is targeting the large cloud providers, but also data center and software solutions providers like WWT to help them provide specific vertical solutions for problems like fraud detection, as well as OEMs and ODMs.

Tanach tells me that the team’s work with Xilinx created the groundwork for its custom chip — though building that (and likely on an advanced node), will cost money, so he’s already thinking about raising the next round of funding for that.

“We are already consuming huge amounts of AI in our day-to-day life and it will continue to grow exponentially over the next five years,” said Tanach. “In order to make AI accessible to every organization, we must build affordable infrastructure that will allow innovators to deploy AI-based applications that cure diseases, improve public safety and enhance education. NeuReality’s technology will support that growth while making the world smarter, cleaner and safer for everyone. The cost of the AI infrastructure and AIaaS will no longer be limiting factors.”

NeuReality team. Photo credit - NeuReality

Image Credits: NeuReality

#artificial-intelligence, #cardumen-capital, #computing, #ethernet, #fpga, #funding, #fundings-exits, #habana-labs, #hardware-startup, #intel, #mellanox, #ml, #neureality, #nvidia, #ourcrowd, #recent-funding, #science-and-technology, #startups, #tc, #technology, #varana-capital, #xilinx


MIT aims to speed up robot movements to match robot thoughts using custom chips

MIT researchers are looking to address the significant gap between how quickly robots can process information (relatively slowly), and how fast they can move (very quickly thanks to modern hardware advances), and they’re using something called ‘robomorphic computing’ to do it. The method, designed by MIT Computer Science and Artificial Intelligence (CSAIL) graduate Dr. Sabrina Neuman, results in custom computer chips that can offer hardware acceleration as a means to faster response times.

Custom-built chips tailored to a very specific purpose are not new – if you’re using a modern iPhone, you have one in that device right now. But they have become more popular as companies and technologists look to do more local computing on devices with more conservative power and computing constraints, rather than round-tripping data to large data centers via network connections.

In this case, the method involves creating hyper-specific chips that are designed based on a robot’s physical layout and and its intended use. By taking into account the requirements a robot has in terms of its perception of its surroundings, its mapping and understanding of its position within those surroundings, and its motion planning resulting from said mapping and its required actions, researchers can design processing chips that greatly increase the efficiency of that last stage by supplementing software algorithms with hardware acceleration.

The classic example of hardware acceleration that most people encounter on a regular basis is a graphics processing unit, or GPU. A GPU is essentially a processor designed specifically for the task of handling graphical computing operations – like display rendering and video playback. GPUs are popular because almost all modern computers run into graphics-intensive applications, but custom chips for a range of different functions have become much more popular lately thanks to the advent of more customizable and efficient small-run chip fabrication techniques.

Here’s a description of how Neuman’s system works specifically in the case of optimizing a hardware chip design for robot control, per MIT News:

The system creates a customized hardware design to best serve a particular robot’s computing needs. The user inputs the parameters of a robot, like its limb layout and how its various joints can move. Neuman’s system translates these physical properties into mathematical matrices. These matrices are “sparse,” meaning they contain many zero values that roughly correspond to movements that are impossible given a robot’s particular anatomy. (Similarly, your arm’s movements are limited because it can only bend at certain joints — it’s not an infinitely pliable spaghetti noodle.)

The system then designs a hardware architecture specialized to run calculations only on the non-zero values in the matrices. The resulting chip design is therefore tailored to maximize efficiency for the robot’s computing needs. And that customization paid off in testing.

Neuman’s team used an field-programmable gate array (FPGA), which is sort of like a midpoint between a fully custom chip and an off-the-shelf CPU, and it achieved significantly better performance than the latter. That means that were you to actually custom manufacture a chip from scratch, you could expect much more significant performance improvements.

Making robots react faster to their environments isn’t just about increase manufacturing speed and efficiency – though it will do that. It’s also about making robots even safer to work with in situations where people are working directly alongside and in collaboration with them. That remains a significant barrier to more widespread use of robotics in everyday life, meaning this research could help unlock the sci-fi future of humans and robots living in integrated harmony.

#artificial-intelligence, #computer-chips, #computer-graphics, #computer-hardware, #computing, #fpga, #hardware-acceleration, #iphone, #mit, #robotics, #science, #tc


The Super-est SuperFX: An unmodified SNES, revved up with ray tracing

If you’ve ever wondered exactly how far a Super Nintendo could be pushed, today’s surprise reveal of a brand-new SNES cartridge hack, as made by a single engineer, is for you. Behold: the SuperRT chip, a proof of concept of how the “SuperFX” idea of the ’90s might have worked with unlimited budgets.

As developed by Ben Carter, an engineer with game-programming credits in game series like Harry Potter, FIFA, and even the 3DS port of Star Fox 64, the SuperRT project delivers pure ray-tracing performance on existing, unmodified SNES hardware. While the SuperRT looks quite unwieldy as a home project, with wires jutting out every which way, you could conceivably slap it into any SNES purchased at a store, then watch it manage real-time light, reflections, and shadows with zero rasterization. It additionally can generate 3D shapes like spheres and planes, then have them intersect in additive fashion to create custom shapes.

The result is a remarkably ’90s-looking CGI demonstration, with circular shapes and planes adding to and subtracting from each other while smothered in large swaths of primary colors. This is all the stuff of intense mathematical calculations, not high-res texture trickery enabled by a glut of VRAM. Yet even without realistic textures or smooth color gradients, the realistic light-bounce results and accurate reflections (including effects like inverted concave mirrors) make the scene look particularly alive.

Read 5 remaining paragraphs | Comments

#fgpas, #fpga, #gaming-culture, #super-nintendo, #superfx


Analogue Pocket’s FPGA-fueled features revealed: $200 pre-order on Aug. 3

After its reveal last October, Analogue‘s portable, retro powerhouse system, the Analogue Pocket, is getting closer to reaching our hands. Today, Analogue has confirmed that Pocket’s pre-order program will kick off Monday, August 3, for $200. That news comes with a delay, however, with the portable system’s original “2020” window being pushed back to May 2021.

As we learned last year, there’s a lot built into this $199 device. The biggest sales pitches include dedicated support for Game Boy, Game Boy Color, and Game Boy Advance cartridges, a “hardware-emulation” backbone as powered by a field-programmable gate array (FPGA) board, and an overkill display resolution of 1600×1440 pixels. Today, Analogue answered most of our remaining questions, and most, but not all, of the answers are good news.

Subpixel gaps, thick glass, and cartridge adapters

First is the screen, which Analogue confirms will ship with a 1.5mm Gorilla Glass covering on its 3.5-inch LTPS LCD display. Though we haven’t gotten exact clarification on the hardware’s in-game color options, particularly for classic monochrome Game Boy games, Analogue is keen to show off its newly announced “original display modes” feature. This takes advantage of the system’s overkill resolution to emulate the subpixel gap inherent in original portable Nintendo hardware, as shown on games for GB, GBC, and GBA, and the sample images thus far look quite handsome. We’ve yet to notice any uneven pixel scaling or other faulty image-scaling issues.

Read 9 remaining paragraphs | Comments

#analogue, #analogue-pocket, #fpga, #gaming-culture


Max Q: Huge week ahead for SpaceX and Virgin Orbit

This week could be the biggest week to date for private spaceflight, with landmark launch attempts coming from both Virgin Orbit and SpaceX .

Virgin Orbit is looking to join the elite club of private launch companies that have actually made it to space, with a full flight test of its combined Cosmic Girl and LauncherOne system. Meanwhile, SpaceX is looking to launch its Crew Dragon spacecraft with people on board – achieving a number of milestones, including returning U.S. crew launch capabilities, and human-rating its Falcon 9 rocket.

Here’s what Virgin Orbit hopes their first flight will do

Virgin Orbit 87Virgin Orbit was supposed to launch its first full demonstration flight on Sunday, but a sensor bug that showed up during pre-launch checkouts means that it’s now pushing things back to at least Monday to check that out.

Extra precaution is hardly surprising since this milestone mission could help the company become an operational satellite launch provider – one of only a small handful of private companies that can make that claim.

SpaceX cleared to proceed for historic crew flight Wednesday

SpaceX passed its first crucial flight readiness review (FRR) on Friday for its first ever crewed astronaut launch, setting it up for a full rehearsal of the mission on Saturday leading up to the actual launch Now it’s set for another FRR with partner NASA on Monday, and then the launch should take place on Wednesday – weather and checkouts permitting. This will definitely be one to watch.

MHI retires a space workhorse


Mitsubishi Heavy Industries flew its last mission with its H-II series rocket, and the space transfer vehicle it carries to deliver supplies to the International Space Station. The company is readying a successor to this highly successful and consistent rocket, the H3, which is set to make its launch debut sometime in 2022 if all goes to plan.

NASA human spaceflight chief abruptly resigns

While SpaceX is aiming to make history with NASA and two of its astronauts, the person in charge of the agency’s human spaceflight endeavors made a surprising and abrupt exit from the agency last week. Doug Loverro resigned from his position, reportedly over some kind of inappropriate activity he engaged in with a prospective agency business partner ahead of the contract awards for NASA’s commercial human lander program.

Xilinx debuts a new chip made for machine learning in space

Xilinx specializes in building processors that are designed to withstand the rigors of use in space, which include heavy radiation exposure, extreme temperatures and plenty more. The company just debuted a new FPGA for space-based applications that is the first 20nm-based processor for space, and the first with dedicated machine-learning capabilities built in for edge computing that truly redefines the term.

NASA’s ‘Artemis Accords’ look to redefine international space cooperation

Space has enjoyed a period of being relatively uncontested when it comes to international squabbles – mostly because it’s hard and expensive to reach, and the benefits of doing so weren’t exactly clear 30 to 40 years ago when most of those rules were set up. NASA’s new rules include a lot of the old ones, but also set up some modernizations that are sure to begin a lot of debate and discussion in the space policy community.

ULA launches first U.S. Space Force spaceplane mission

100330 O 1234S 001

In a testing procedure, the X-37B Orbital Test Vehicle taxis on the flightline March 30, 2010, at the Astrotech facility in Titusville, FLa. (Courtesy photo)

The United Launch Alliance launched the X-37B last week on behalf of the U.S. Space Force – marking the first time the mysterious experimental unscrewed space plane has launched for that newly-formed agency. The X-37B has flown plenty before, of course – but previously it was doing so under the authority of the U.S. Air Force, since the Space Force hadn’t been formed yet.

#aerospace, #astronaut, #computing, #doug-loverro, #falcon, #falcon-9, #flight, #florida, #fpga, #international-space-station, #machine-learning, #outer-space, #private-spaceflight, #radiation, #space, #spaceflight, #spaceplane, #spacex, #tc, #u-s-air-force, #united-states, #xilinx


Xilinx launches a new reconfigurable space-grade chip optimized for local machine learning on orbit

Space-specific silicon company Xilinx has developed a new processor for in-space and satellite applications that records a number of firsts: It’s the first 20nm process that’s rated for use in space, offering power and efficiency benefits, and it’s the first to offer specific support for high performance machine learning through neural network-based inference acceleration.

The processor is a field programmable gate array (FPGA), meaning that customers can tweak the hardware to suit their specific needs since the chip is essentially user-configurable hardware. On the machine learning side, Xilinx says that the new processor will offer up to 5.7 tera operations per second of “peak INT8 performance optimized for deep learning,” which is an improvement of as much as 25x vs the previous generation.

Xilinx’s new chip has a lot of potential for the satellite market for a couple of reasons: First, it’s a huge leap in terms of processor size, since the company’s existing traditional tolerant silicon was offered in a 65nm spec only. That means big improvements in terms of its size, weight and power efficiency, all of which translates to very important savings when you’re talking about in-space applications, since satellites are designed to be as lightweight and compact as possible to help defray launch costs and in-space propellant needs, both of which represent major expenses in their operation.

Finally, its reconfigurable nature means that on-orbit assets can be reprogrammed on-demand to handle different tasks – which now include local machine learning algorithm processing. That means you could theoretically switch one of these in an Earth observation satellite from handling something like tracking cloud density and weather patterns, to making inferences about deforestation or strip mining, for instance. That’s a whole lot of added flexibility for satellite constellation operators looking to move where market demand is needed most.

Xilinx’s chips are special in a number of ways vs. the kind we use here on Earth, including with the aforementioned radiation tolerance. They also come packed in thick ceramic packaging which add extra durability both during the launch phase, where stresses include extreme vibration, and on orbit where the lack of an atmosphere means exposure to an extremely harsh environment in terms f both radiation and temperature.

#aerospace, #fpga, #machine-learning, #neural-network, #radiation, #satellite, #science, #semiconductors, #space, #tc, #technology, #xilinx