Embedding tables dominate industrial-scale recommendation model sizes, using up to terabytes of memory. A popular and the largest publicly available machine learning MLPerf benchmark on recommendation data is a Deep Learning Recommendation Model (DLRM) trained on a terabyte of click-through data. It contains 100GB of embedding memory (25+Billion parameters). DLRMs, due to their sheer size and the associated volume of data, face difficulty in training, deploying for inference, and memory bottlenecks due to large embedding tables. This paper analyzes and extensively evaluates a generic parameter-sharing setup (PSS) for compressing DLRM models.

Micron has shipped the first PCI Express 6.0 SSD, ramping up read and write speeds to unprecedented levels. This week, Micron shipped the Micron 9650 SSD, the world's first PCIe 6.0 SSD, designed for AI training and inference workloads. Unfortunately, those tasks take place in AI data centers, not home PCs. Micron will ship the drive in both a PRO (read-intensive) and MAX (write-intensive) configuration, with capacities ranging from 6.4TB to 30.72TB, depending on which flavor a customer buys. Technically, the drives use a PCI Express 6.2 interface, connecting to Micron's six-plane, ninth-generation (G9) flash memory.

Google on Tuesday unleashed another wave of artificial intelligence technology to accelerate a year-long makeover of its search engine that is changing the way people get information and curtailing the flow of internet traffic to other websites. The next phase outlined at Google's annual developers conference includes releasing a new "AI mode" option in the United States. The company says the feature will make interacting with its search engine more like having a conversation with an expert capable of answering a wide array of questions. AI mode is being offered to all users in the US just two-and-a-half months after the company began testing with a limited Labs division audience. Google is also feeding its latest AI model, Gemini 2.5, into its search algorithms and will soon begin testing other AI features, such as the ability to automatically buy concert tickets and conduct searches through live video feeds.

Embedding tables dominate industrial-scale recommendation model sizes, using up to terabytes of memory. A popular and the largest publicly available machine learning MLPerf benchmark on recommendation data is a Deep Learning Recommendation Model (DLRM) trained on a terabyte of click-through data. It contains 100GB of embedding memory (25 Billion parameters). DLRMs, due to their sheer size and the associated volume of data, face difficulty in training, deploying for inference, and memory bottlenecks due to large embedding tables. This paper analyzes and extensively evaluates a generic parameter-sharing setup (PSS) for compressing DLRM models.

Diamonds aren't just a luxury item--as one of the hardest naturally occurring materials in existence, they are vital components in many industrial drills, medical devices, and even space-grade materials. But recent scientific advancements show it's not just their durability that's impressive, but their data storage capabilities. According to a study published on November 27th in the journal Nature Photonics, researchers at China's University of Science and Technology in Hefei have achieved a record-breaking diamond storage density of 1.85 terabytes per cubic centimeter. CDs, solid state drives, and Blu-ray discs are well suited to handle most general data storage needs, but that increasingly isn't the case for projects requiring massive amounts of digitized information. The artificial intelligence industry as well as quantum and supercomputers often need petabytes, not gigabytes or even terabytes, of information storage.

This week scientists published the highest-resolution map yet of one small piece of the brain, a tissue sample one cubic millimeter in size. The resulting data set comprised 1,400 terabytes. This map is just one of many that have been in the news in recent years. So this week I thought we could walk through some of the ways researchers make these maps and how they hope to use them. Scientists have been trying to map the brain for as long as they've been studying it.

NVIDIA has launched its next-generation of AI supercomputer chips that will likely play a large role in future breakthroughs in deep learning and large language models (LLMs) like OpenAI's GPT-4, the company announced. The technology represents a significant leap over the last generation and is poised to be used in data centers and supercomputers -- working on tasks like weather and climate prediction, drug discovery, quantum computing and more. The key product is the HGX H200 GPU based on NVIDIA's "Hopper" architecture, a replacement for the popular H100 GPU. It's the company's first chip to use HBM3e memory that's faster and has more capacity, thus making it better suited for large language models. "With HBM3e, the NVIDIA H200 delivers 141GB of memory at 4.8 terabytes per second, nearly double the capacity and 2.4x more bandwidth compared with its predecessor, the NVIDIA A100," the company wrote.

More compute leads to greater electricity consumption, and consequent carbon emissions. A 2019 study by researchers at the University of Massachusetts Amherst estimated that the electricity consumed during the training of a transformer, a type of deep learning algorithm, can emit more than 626,000 pounds ( 284 metric tons) of carbon dioxide--equal to more than 41 round-trip flights between New York City and Sydney, Australia. We are also facing an explosion of data storage. IDC projects that 180 zettabytes of data--or, 180 billion terabytes--will be created in 2025. The collective energy required for data storage at this scale is enormous and will be challenging to address sustainably.

There is no doubt that GPT-4, the latest iteration of the artificial-intelligence engine created by the company OpenAI, is innovative and cool. It can create a poem in the style of Basho, spell out the chord progression and time signature for a simple tune, and provide a seven-step recipe for a peanut-butter-and-jelly sandwich. When I asked it to write a musical about a narcissistic politician who holds the fate of the world in his hands, it delivered a story in two acts, with a protagonist named Alex Sterling who "navigates a maze of power, manipulation, and the consequences of his decisions," as he sings "Narcissus in the Mirror," "The Price of Power," and about a dozen other invented songs. Those songs appear to have been created out of thin air; certainly, no human conceived them. Still, Alex's story, which "explores themes of self-discovery, redemption, and the responsibility of leadership," is quite familiar.

BOCA RATON, Fla., Nov. 15, 2022 (GLOBE NEWSWIRE) -- Totaligent, Inc. ("Totaligent" or "the Company") (OTCPK: TGNT) announces it has completed testing of its scalable Nvidia clusters and has started to build a super cluster, with 2.4 Terabytes of GPU ram and 18 Terabytes of system ram. Totaligent's new supercomputer will allow the Company's Artificial Intelligence to deliver nearly instantaneous data processing and modeling for its person-based digital marketing platform. "Having the power and speed to deliver near real-time results when building target audiences from billions of records for customers is critical to Totaligent's success and acceptance in the person-based digital marketing world. Now, when we append large datasets that used to take days to process, our AI completes the task in about a minute. The combination of data, speed, and a complete set of easy-to-use marketing tools, at an affordable price, will enable Totaligent to provide unparalleled results for its users upon the launch of its integrated digital platform," stated Ted DeFeudis, CEO.