The European Commission will soon present its European Chips Act, a plan to reduce supply chain dependencies on chip manufacturers outside the bloc. The plan was detailed by EC president Ursula von der Leyen at the 2021 State of the Union address. It's become standard for mid-range and flagship phones across all major US wireless carriers and most new phones incorporate 5G technology. "The aim is to jointly create a state-of-the-art European chip ecosystem, including production," she said. "That ensures our security of supply and will develop new markets for ground-breaking European tech."
An IBM researcher holds a silicon wafer with embedded IBM Telum chips designed to maximize artificial intelligence capabilities. The chips were developed at Albany Nanotech and made in partnership with Samsung. The Albany area was recent cited by the Brookings Institution for having the potential to create an AI sector. ALBANY -- The Capital Region is one of 87 "potential adoption centers" in the United States for companies and researchers focused on the use of artificial intelligence, or AI, according to a new report from the Brookings Institution, a left-leaning think tank. The San Francisco Bay area is No. 1 in AI, while other upstate cities, Buffalo, Rochester and Syracuse, were also listed as potential adoption centers.
Sony issued a disclaimer ahead of Thursday's showcase that the next generation of PlayStation's virtual reality tech wouldn't make an appearance. Also absent was a much-anticipated update about the PS5′s expanded storage feature. Over the summer, Sony launched a beta program for the long-awaited feature, which allows users to expand their device's storage using specific M. 2 solid-state drives, for select users in the United States, Canada, France, Japan, Germany and the U.K. When it will roll out to the general public remains unclear.
Inside a large clean room in rural Connecticut, engineers have begun constructing a critical component for a machine that promises to keep the tech industry as we know it on track for at least another decade. The machine is being built by ASML, a Dutch company that has cornered the market for etching the tiniest nanoscopic features into microchips with light. ASML introduced the first extreme ultraviolet (EUV) lithography machines for mass production in 2017, after decades spent mastering the technique. The machines perform a crucial role in the chipmaking ecosystem, and they have been used in the manufacture of the latest, most advanced chips, including those in new iPhones as well as computers used for artificial intelligence. The company's next EUV system, a part of which is being built in Wilton, Connecticut, will use a new trick to minimize the wavelength of light it uses--shrinking the size of features on the resulting chips and boosting their performance--more than ever before.
Long Short-Term Memory (LSTM) recurrent networks are frequently used for tasks involving time-sequential data such as speech recognition. However, it is difficult to deploy these networks on hardware to achieve high throughput and low latency because the fully connected structure makes LSTM networks a memory-bounded algorithm. Previous LSTM accelerators either exploited weight spatial sparsity or temporal activation sparsity. This paper proposes a new accelerator called "Spartus" that exploits spatio-temporal sparsity to achieve ultra-low latency inference. The spatial sparsity is induced using our proposed pruning method called Column-Balanced Targeted Dropout (CBTD), which structures sparse weight matrices for balanced workload. It achieved up to 96% weight sparsity with negligible accuracy difference for an LSTM network trained on a TIMIT phone recognition task. To induce temporal sparsity in LSTM, we create the DeltaLSTM by extending the previous DeltaGRU method to the LSTM network. This combined sparsity simultaneously saves on the weight memory access and associated arithmetic operations. Spartus was implemented on a Xilinx Zynq-7100 FPGA. The Spartus per-sample latency for a single DeltaLSTM layer of 1024 neurons averages 1 us. Spartus achieved 9.4 TOp/s effective batch-1 throughput and 1.1 TOp/J energy efficiency, which, respectively, are 4X and 7X higher than the previous state-of-the-art.
Qualcomm launched a drone platform that aims to replicate the success of the Ingenuity Helicopter on Mars on Earth. The company's technology and its Qualcomm Flight Platform was used for the unmanned flight of the Ingenuity Helicopter. That effort turned out to be a test for the Qualcomm Flight RB5 5G Platform, which is a reference design for drones that integrates 5G and AI. Qualcomm's Flight RB5 5G Platform aims to accelerate development for commercial, enterprise and industrial drones as well as edge computing. Target use cases include entertainment, security, delivery, defense, inspection and mapping.
Wan, Weier, Kubendran, Rajkumar, Schaefer, Clemens, Eryilmaz, S. Burc, Zhang, Wenqiang, Wu, Dabin, Deiss, Stephen, Raina, Priyanka, Qian, He, Gao, Bin, Joshi, Siddharth, Wu, Huaqiang, Wong, H. -S. Philip, Cauwenberghs, Gert
Realizing today's cloud-level artificial intelligence functionalities directly on devices distributed at the edge of the internet calls for edge hardware capable of processing multiple modalities of sensory data (e.g. video, audio) at unprecedented energy-efficiency. AI hardware architectures today cannot meet the demand due to a fundamental "memory wall": data movement between separate compute and memory units consumes large energy and incurs long latency. Resistive random-access memory (RRAM) based compute-in-memory (CIM) architectures promise to bring orders of magnitude energy-efficiency improvement by performing computation directly within memory. However, conventional approaches to CIM hardware design limit its functional flexibility necessary for processing diverse AI workloads, and must overcome hardware imperfections that degrade inference accuracy. Such trade-offs between efficiency, versatility and accuracy cannot be addressed by isolated improvements on any single level of the design. By co-optimizing across all hierarchies of the design from algorithms and architecture to circuits and devices, we present NeuRRAM - the first multimodal edge AI chip using RRAM CIM to simultaneously deliver a high degree of versatility for diverse model architectures, record energy-efficiency $5\times$ - $8\times$ better than prior art across various computational bit-precisions, and inference accuracy comparable to software models with 4-bit weights on all measured standard AI benchmarks including accuracy of 99.0% on MNIST and 85.7% on CIFAR-10 image classification, 84.7% accuracy on Google speech command recognition, and a 70% reduction in image reconstruction error on a Bayesian image recovery task. This work paves a way towards building highly efficient and reconfigurable edge AI hardware platforms for the more demanding and heterogeneous AI applications of the future.
We introduce OPtical ADversarial attack (OPAD). OPAD is an adversarial attack in the physical space aiming to fool image classifiers without physically touching the objects (e.g., moving or painting the objects). The principle of OPAD is to use structured illumination to alter the appearance of the target objects. The system consists of a low-cost projector, a camera, and a computer. The challenge of the problem is the non-linearity of the radiometric response of the projector and the spatially varying spectral response of the scene. Attacks generated in a conventional approach do not work in this setting unless they are calibrated to compensate for such a projector-camera model. The proposed solution incorporates the projector-camera model into the adversarial attack optimization, where a new attack formulation is derived. Experimental results prove the validity of the solution. It is demonstrated that OPAD can optically attack a real 3D object in the presence of background lighting for white-box, black-box, targeted, and untargeted attacks. Theoretical analysis is presented to quantify the fundamental performance limit of the system.
When someone mentioned to me that Xiaomi was launching its own "robot dog," my mind immediately went to Sony's Aibo. And honestly, it would have been difficult to be more wrong. Now that the news has been out for a few days, the company's heard all of your bad Black Mirror jokes, don't worry. And, honestly, the Chinese hardware maker didn't do itself any favors with the design here. Boston Dynamics has done a lot to imbue its quadrupedal robots with personality, through design language and viral videos of Spot and company busting a move to the Dirty Dancing soundtrack.
"We're seeing 10- to 12-week delivery times for laptops and computing devices," said Sue Workman, chief information officer at Case Western Reserve University in Cleveland. "Those used to take a day or two." For the rapidly approaching fall semester, the school is hustling to equip classrooms with video displays, microphones and other tools so students have the option of taking some classes from home, Ms. Workman said. Orders for both displays and microphones have been delayed, she added. The Morning Download delivers daily insights and news on business technology from the CIO Journal team.